Document 12300144

advertisement
Table of Contents
The applicability of the IS-PC-MR model on the UK economy during the crisis by
Jia Wei Mark Choo ........................................................................................................................... 7
On target: an analysis of Abenomics by Chen Yue Lok ..................................................... 16
The Value of Theoretical Models for Real-World Market Analysis by Maryam Khan. 24
A Dynamic, Hierarchical, Bayesian approach to Forecasting the 2014 US Senate
Elections by Roberto Cerina ...................................................................................................... 30
Arming Women With Credit: Why Lending to Women Makes Economic Sense by
Victoria Monro ............................................................................................................................... 46
Effect of the appreciation of the Swiss franc on the Ticinian Job Market by Guido
Tubaldi .............................................................................................................................................. 54
Localised complementary currencies: the new tool for policymakers? The Sardex
exchange system............................................................................................................................... 66
Savings-Linked, Saving Lives: The Role of Conditional Cash Transfers in Financial
Inclusion by Isaac Lim ................................................................................................................. 80
Breaking up Germany: A review of the drivers of social mobility by Dennis
Dinkelmeyer ................................................................................................................................... 92
The effects of television viewing on income aspirations and happiness by Emma
Claydon .......................................................................................................................................... 104
Bringing the Vanguard to the Rearguard: A transformative development model
for Brazil’s backlands By James B Theuerkauf * ............................................................. 116
Addressing the Exchange Rate Puzzle: Forecasting with a Taylor rule inspired
model by Amy Louise Curry.................................................................................................... 131
Poster presentations................................................................................................................. 144
Acknowledgements ................................................................................................................... 144
3
Session 1: Exploring the value of
economic models
5
The applicability of the IS-PC-MR model on the UK economy during
the crisis by Jia Wei Mark Choo
2nd year, UCL Economics
Introduction
Since the 2008 great recession, it has been a very interesting time for macroeconomics. There has
been a movement to revamp how economics is taught, challenging the ability of models and
mathematical equations to map agents’ behaviour and the economy. This debate has inspired my
interest on the applicability of economic theories. It is argued that oversimplification and
generalization of assumptions make models too abstract and impractical, yet, at various points in time,
models have been relatively successful in mapping out general movements in economies. As logical
beings, we are constantly seeking coherence in daily occurrences and given all its drawbacks,
economic models might still currently be the best solution to understanding reality. A core model of
macroeconomics is the IS-PC-MR model (3EM), and having just spent a good portion of time
learning the closed economy 3EM, I thought it was apt to examine if this model accurately reflects
reality and its ability to analyse shocks of such magnitude.
The paper will firstly set up the 3EM, following that, it will apply the model onto the UK closed
economy over 2008. This is done by analysing the Bank of England (BoE) inflation report, drawing
out relevant information specific to the UK domestic economy and applying it to the model. This
paper uses those reports as they are a reliable source of information and they provide the necessary
data for the analysis. I will be picking up key phrases from the BoE report overview in every period
and use it to explain the intuition behind the movements of the curves in the analysis and map out
developments in the economy. After that, I will compare and review the proximity between the
model’s prediction and what actually happened, followed by drawing conclusions about the
applicability of models.
IS-PC-LM Model
The 3EM is a macroeconomic model which incorporates the demand and the supply side of the
economy, and the (CB) Bank’s Monetary Rule at any given period.
IS
The IS curve reflects the demand side of the economy and in the closed economy and it is given as
Consumption + Investment + Government Spending. The components of Consumption and
Investment are negatively related to real interest rate set by the CB. Government Spending is
autonomous.
PC
The Phillips Curve (PC) reflects the supply side of economy. It is generated from the wage- and pricesetting (WS/PS) behaviour of the economy. As workers bargain for real wages and firms setting
prices, the PC is the relation between inflation and output in the economy, where it is a feasible set of
inflation and output pairs for a given rate of expected inflation p .
e
7
MR
The Monetary Rule (MR) curve reflects the best-response behavior of the inflation-targeting central
bank and it is the optimum combinations of output and inflation that the CB will choose subject to the
PC it faces. The Bank of England aims to deliver price stability, growth and employment, i.e. keeping
inflation constant at its target rate of πT and output at its max potential ye in the medium run
equilibrium (MRE). The MR generates the real interest rate (Bank Rate) the CB needs to set to guide
the demand side of the economy towards its target.
3EM analysis against BoE reports
Before I begin, I shall lay out the scope of my analysis. Firstly, I assumed that the UK economy is a
closed economy and assumed away trade, exchange rate and the forex market. This was done not
because the UK economy resembles a closed economy, but for a simplification reason. The aim of the
paper was to observe the newly learnt closed-economy model in practice and thus, the UK was simply
assumed to be purely domestic.
This paper analyses the UK economy specifically in 2008 because it was when the great recession
was unfolding and changing interest rates was still the main tool the Bank of England (BoE) used.
Other tools such as Quantitative Easing started in 2009 and thus, periods after 2008 were left out of
the analysis.
The length of each period of analysis depends on “stickiness” of wages. It will be assumed that the
wage-setting period has a length of three months, in line with the BoE quarterly inflation reports.
Key Quotes from BoE Inflation Report
February 2008
Central projection
“output growth to slow markedly this year and
gradually start to recover” (pg 5)
“pay growth was steady” (ibid)
“some measures of inflation expectations rose” (ibid)
“Higher energy, food and import prices push
inflation up sharply in the near term” (ibid)
“consumer spending growth appeared to soften and
the climate for investment deteriorated” (ibid)
“business surveys and reports from the Bank’s
regional Agents point to a further modest
deceleration in activity in early 2008” (pg 6)
“the extent to which consumer prices increase will
depend on whether businesses and retailers can pass
on higher input costs.” (pg 7)
“Survey measures of businesses intend to pass on
cost increases. But there are indications that retailers
have been accepting lower profit margins in order to
maintain sales volumes” (ibid)
8
Going into period 1, “CPI inflation remained close to the 2% target in December” (BoE, Feb 2008: 5)
at 2.1% and “output growth moderated to around its long-term historical average rate” (ibid). It will
be assumed that the economy is at MRE point z, where central bank is on its target with inflation at
2% and output at ye.
Assuming the great recession only started in the beginning of 2008, the UK economy is hit by a
negative demand shock in period 1. IS curve shifts from IS to IS' and given the current bank rate of
5.50%, The economy moves from point z to point A, with output below ye at y1'.
However, the economy is also experiencing an oil/commodity price rise. CB is away from its bliss
point and forecasts next period’s PC to be at PC(πe=π1’), at a higher inflation rate. Faced with this PC,
CB would like to be on point B on its MR curve with output at y1 and inflation at π1. In order to
achieve this, CB sets r such that when corresponding to IS’, it will result in a lower bank rate than the
current rate. As the fall in domestic demand is forecast to be “modest” and with inflation expectations
rising, the CB has to “balance conflicting risks” and reduced Bank Rate only slightly. This is reflected
in its policy rate falling from 5.50% to 5.25% in its February 2008 Inflation Report.
The oil/energy price increase is analysed as a PC inflation shock instead of modeled by the change in
markup, µ, of the Price-Setting curve because of BoE’s believe that “inflation [will ease] back to a
little above the 2% target in the medium term, as the near-term rise in energy prices drops out of the
twelve-month rate” (ibid: 7). If this shock is interpreted as a supply side disturbance, ye will change
and bank's mandate (ie the MR curve) will change as well. However, this is not the case as the BoE
projects the medium term output to be back to its original level.
Key Quotes from BoE Inflation Report
May 2008
Central projection
“output growth slows markedly through 2008, opening up
a margin of spare capacity. Sluggish income growth and
tighter credit supply dampen consumer spending, while
weaker demand outlook and lower property prices also
weigh on business and residential investment.” (pg 7)
“CPI inflation was 2.5% in March” (pg 5)
“higher energy and import prices push inflation up sharply
in the near term” (ibid)
“Pay growth remained muted” (ibid)
“measures of short-term household inflation expectations
rose again, broadly in line with the recent and prospective
movements in consumer price inflation.” (pg 7)
“The outlook is somewhat weaker than in the February
Report for the first part of the projection” (ibid)
“There were particular uncertainties relating to the
severity of the slowdown and the future path of inflation
expectations.” (pg 8)
In period 2, the forecasted PC(πe=π1’) curve actually happens. This is reflected by an increase in
inflation despite a fall in output as predicted in Feb's report, given bank rate of 5.25%. Economy is
now at point B as predicted.
9
The analysis of the oil/commodity price increase as a PC inflation shock instead of a supply side
shock in period 1 is in accordance with the economic data reflected in the May Report and is justified.
The BoE backs this up by continuing to project that “declining contribution from energy and import
prices, then bring inflation back to around the 2% target in the medium term” (BoE, May 2008: 5).
The negative demand shock continues to persist and demand continues to fall in period 2. The IS
curve shifts leftwards from IS' to IS'' and given the bank rate of 5.25%, there would be a fall in output
from y1 to y2'. This corresponds to point C'.
The high oil and energy prices continue to rise sharply and this leads to the inflation shock to continue
in the next period. CB forecasts next period's PC to be PC(πe=π2’) and faced with this PC, CB would
like to be on point C with output y2 and inflation rate at point π2. This leads to the BoE setting bank
rate to be 5.00%.
Key Quotes from BoE Inflation Report
August 2008
Central projection
“output to be broadly flat over the next year or so, after
which growth gradually recovers”
“Higher energy, food and import prices push inflation
substantially higher over the next few months” (pg 5)
“considerable uncertainty surrounds this outlook”
(ibid)
“output growth eased in the second quarter and surveys
pointed to further slowing in the third” (ibid)
“consumer spending appeared to decelerate as
households’ real incomes were squeezed” and “outlook
for investment deteriorated” (ibid)
“Subsequent data … suggested that growth in Q2 may
have been weaker than provisionally estimated” (pg
6)
“CPI inflation increased to 3.8% in June, reflecting
sharp increases in food and petrol prices” (pg 7)
“companies will over time have to pass the burden of
higher non-labour costs on to households. That will
occur through some combination of lower nominal
wage growth, higher prices and lower employment
growth, implying a reduction in the purchasing power
of households’ incomes.” (ibid)
“CPI inflation is expected to pick up sharply over the
next few months” (ibid)
“Rise in near-term inflation expectations is consistent
with households expecting the rise in inflation to be
temporary” (ibid)
10
In period 3, the forecasted PC(πe=π2’) actually happens. Economy is indeed on point C with output
falling to y2 in previous period and CPI inflation up to 3.8%. Again, the modeling of the oil/energy
price hike via PC inflation shock is justified as in period 2. This is backed up again by the BoE’s
expectation that food and energy prices will fall back in the medium term.
The negative demand shock is a persistent one and continues in period 3. IS curve continues to shift
leftward from IS'' to IS''' and given bank rate at 5.00%, output falls from y2 to y3, from point C to
point D.
Oil/energy prices continue to rise sharply, exacerbating the inflation shock. PC continues to rise to
PC(πe=π3’). Faced with this PC, CB's best response is to be on point D where output is at y3 and
inflation at π3. Given the high levels of inflation, despite the fall in output, CB decides to leave bank
rate unchanged as its best response, guiding the economy to point D.
Key Quotes from BoE Inflation Report
November 2008
Central projection
“Output to continue to fall in the early part of the
forecast, followed by a gradual recovery”
“inflation slows sharply in the near term, as the
contributions from energy and food prices decline”
(pg 5)
Due to fall in outlook for inflation, “committee
judged that a significant reduction in Bank Rate
was necessary in order to meet the 2% target in the
medium term” (pg 8)
“CPI inflation rose to 5.2% in September” (pg 5)
“Near-term outlook for inflation improved
significantly in the wake of sharp falls in
commodity prices” (pg 5)
“Measures of household inflation expectations fell
back and earnings growth remained contained”
(ibid)
“The outlook for domestic demand deteriorated
markedly” and “prospects for investment
worsened” (pg 6)
“key drivers are a sharp tightening in the supply of
money and credit, subdued growth in incomes and
past falls in asset prices” (pg 7)
“The prospects for economic growth and inflation
are judged to be unusually uncertain, reflecting
the exceptional economic and financial factors
affecting the outlook.” (pg 8)
11
In period 4, the forecasted PC(πe=π3’) actually happens as inflation spiked up to 5.2% and output
continued to fall as predicted in Aug report. Economy is on point D. Negative demand shock
continues to persist with IS shifting leftward from IS''' to IS''''. Given rate of 5.00%, this corresponds
to a fall in output to y4' on point E'.
Oil/energy prices have fallen and similar to the previous periods, it will be interpreted with a fall in
inflation expectations. PC shifts down to PC(πe=π4’). As inflation expectations have eased, CB is able
to pursue higher output without compromising its inflation aims. CB reduced bank rate strongly by 2
percentage points to 3.00%, aiming to increase output to y4 and a lower inflation rate to π4.
However, looking at the February 2009 inflation report, the UK economy did not behave as the 3EM
predicted. Though “CPI inflation fell to 3.1% in December” (BoE, February 2009: 5), as rightly
predicted from the fall in PC, “GDP contracted sharply in the fourth quarter of 2008” (ibid), despite a
lowering of the bank rate. This deviation could possibly be explained by the UK economy
experiencing “a substantially larger (GDP) decline than envisaged at the time of the November
Report” (ibid: 6). The IS curve shifted leftward more than the BoE expected and the ‘actual’ IS curve
is at IS’’’’a instead of IS’’’’. At bank rate of 3%, the economy is on point Ea rather than E, with output
falling to y4a and inflation falling to π4a. This is now more consistent with the data of Q4 2008 as
reported in the February 2009 report.
Analysis results compared against reality
As we can see, especially in the first 3 periods, the 3EM correctly modeled the general directions of
inflation, output and Bank Rate and the results were somewhat congruent with the data from the BOE
reports. This gives us a good intuition behind the interaction between the IS, PC and MR curves.
However, complication started in period 4, especially when the crisis was reaching its zenith. In the
November 2008 report, the BoE expected “output to continue to fall”, yet my analysis via the models
did not accurately account for this fall, but even predicted a rise in output. With extra data from
February 2009 report, I attempted to correct my analysis and was slightly more successful in
representing the data.
Possible reasons for differences
Some of the possible reasons for the deviation of my analysis of the 3EM and the information of the
UK economy from the BoE statements will be address below.
Bank of England’s uncertainties
Firstly, the BoE itself is unsure of the situation of the economy. There are lags and measurement
uncertainties in collecting economic data, especially for GDP. In order to project inflation, output and
accordingly set the bank rate in the next period, it is ideal that the CB has perfect information about
the current situation of the economy and future developments. The model assumes that the CB is an
omniscient agent, however, this is far from the case in reality. Though the BoE has rightly predicted a
slowdown, they were unable to predict the extent of the slowdown. Also, there was uncertainty over
the permanency of the negative demand shock. This uncertainty is clearly evident in BoE’s method of
projecting the future of the UK economy via fan diagrams.
The BoE also cannot truly interpret all the economic information and data they have gathered. Given
the complexity of the economy and the limited ability of the Monetary Policy Committee, shocks
cannot be accurately identified and explained. This is clear in the case of the energy prices shock that
accompanied the negative demand shock. The BoE was unsure if firms were going to pass on the
increase in import costs to higher prices, lowering wages or retrenching workers. This makes mapping
12
the energy price shock, as a simple inflation shock, or a supply side shock, or a combination of both,
extremely difficult.
Problems with the 3EM
Firstly, some of the assumptions embedded in the 3EM are not representative of reality. (1) The 3EM
assumes that agents in the demand and supply side of the economy are homogenous. However, this is
not the case as every agent has different preferences and behaviours. (2) The PC is assumed to have
adaptive expectations, while in reality, it tends to be a mix between rational and adaptive
expectations. (3) In order for the CB to be able to have an impact on the economy, it has to be
credible for the demand and supply side of the economy to react to its actions. The BoE is simply
assumed to be credible.
Secondly, the 3EM fails to account for the financial sector, which played an instrumental role in 2008
great recession. The 3EM ignores “characteristics of the financial system that can create vulnerability
to a financial crisis, with implications for fiscal balance” (Carlin and Soskice, 2015: xii). In order to
better understand the global economy, the financial system has to be integrated into macroeconomic
models. Some of such models are proposed in the textbook Macroeconomics: Institutions, Instability,
and the Financial System by Wendy Carlin and David Soskice.
Thirdly, monetary policy in practice is far more complex than the CB setting real interest rates as the
driving force to guide the economy. In practice, other factors such as asset prices, expectations and
confidence, and exchange rates affect Monetary Policy (Carlin and Soskice, 2015: 479). The 3EM
also assumes the Bank Rate is the only method of affecting output, but there are classical mechanisms
or market clearing mechanisms that will put pressure on output. These pressures affect the economy
naturally when output is in disequilibrium. In the 3EM, this is assumed away.
Another possible reason for the deviation between the projection of the BoE and reality attributable to
the model is the assumption that the real rate set by the CB is the rate of credit of the economy.
However, this is not the case as BoE Bank Rate and actual lending rates, such as LIBOR, were not
identical. Even the BoE reports acknowledge this drawback and projects its view on output and
inflation “assuming the bank rate follows market yields”.
When using the 3EM, the economy is viewed as static period blocks. Economic data in every period
is analysed as changes in absolute value and block shifting of IS and/or PC curves. In reality, this is
hardly the case as fluid movements and constant gradual shifting is a better representation of the
behaviour of the economy.
Another setback of the 3EM is that it can only analyse single shocks at a time. When multiple shocks
hit the economy in reality, the 3EM cannot accurately analyse changes to inflation and output.
Multiple shocks tend to come hand in hand and when these shocks affect economic variables in
opposite direction, the simple 3EM is unable to discern which shock outweighs which or generate
accurately the final outcome.
Limitations to my methodology
Another possible explanation for the difference between the model analysis and reality is the
existence of limitations in the process of my analysis.
Firstly, I assumed that the UK economy is a closed economy. However, the UK is extremely open and
globalised, it is not good representation of a closed economy. Factors such as trade, exchange rates
and the forex market are needed to be accountable in order to have a more accurate interpretation of
the UK economy.
13
Secondly, I assumed that the UK economy initially started at equilibrium output ye and that it has
reached its max potential, ie that it is not growing anymore. Thus, I treated the slowing of output
growth in the BoE statements as comparable to a fall in output. This poses 2 problems. (1) I have
inaccurately used the model right from the beginning. Instead of modeling a growing economy with
intrinsically non-static IS, PC and MR curves, I assumed that the economy is static at MRE for
simplification. (2) The UK economy might not have been at ye at the start of 2008. The issue of
deciding what exactly is “equilibrium output” limits the accuracy of my analysis.
Thirdly, when analysing the economic data from the reports onto the model, I was neither pedantic
nor technical in taking into account the exact change in output, inflation and bank rate. Economic
information were merely modeled as direction shifting on the 3 curves and not meticulously drawn to
scale. This lack of accuracy on my part possibly allowed for a substantial margin of error for my
model analysis.
Fourthly, as this paper is analysing past events, there is a possibility of hindsight bias in my modeling.
I was presented with an ‘end point’ to work towards and that might have blinded my analysis. Instead
of simulating present time economic data analysis, I might have fallen into the trap of forcing the
model to fit the data.
Fifthly, there are many ways to interpret the same economic data and therefore my interpretation is
neither the only nor the ‘right’ interpretation. Even the Monetary Policy Committee themselves are
using a variation of the 3EM as a tool! An example is mapping the oil/energy price shock as a fullfledged PC inflation shock. As pointed out under the section of BoE’s uncertainty, I do not know the
extent of the shock as a temporary PC inflation shock or having deeper supply side repercussions.
Others might interpret such data differently and will ultimately get very different results.
Strengths and limitations of modeling
After observing a core macroeconomic model in practice, we can now look at the strengths and
limitations of modeling in economics.
Strengths
It is argued that economic models can help us understand the mechanisms of economics and in the
case of the 3EM, macroeconomics. The 3EM assists us in understanding the effects of individual
policies of the CB for the next periods given the best information we have on what has occurred in
practice. As such, models help us analyse a very complex world by breaking down data into sections
and add clarity to that understanding.
Limitations
On the other hand, it is argued that as the global economy is extremely complex and with economic
models being extremely simple in comparison, results from modeling is inaccurate and unreliable.
The unrepresentative assumptions and oversimplifications possibly make models impractical and
obsolete. Many shocks and developments were bombarding the economy during the great recession
and the 3EM might be too simplified to handle such developments.
Mathematics, the tenet of economic models, tends to create general equations, unnecessarily
standardising economic agents and their behaviour. In such equations, parameters and variables are
given values within models. However, the true values of such parameters are unknown and possible
constantly changing! It is argued that economics should not be treated this way as the generalisations
destroy details which are pertinent to social sciences. This is because in reality, there exists so many
14
different actors with multiple fast-moving developments happening at the same time, it is impossible
to trace out what exactly is happening in the economy.
From this paper’s analysis, we can see a glimpse of the workability of models. A model is analogous
to a ‘machine’, churning out information based on the ‘settings’ input. Though it is obvious that it
cannot exactly predict the outcome of reality per se, based on the information that is put into the
machine of the model, it does however reflect quite accurately what happens and/or what can happen.
Models are effective in mapping predictions BUT NOT PREDICT. This is evident in the analysis of
period 4 where after the addition of the data from the February 2009 report, the model produced a
closer match to reality.
Conclusion
“Essentially, all models are wrong, but some are useful” (Box, 1987: 424). Though models might be
an unrepresentative reflection of reality with all its illogical assumptions and oversimplifications,
models do indeed help us map out our thoughts and give structure to our analysis of economics. The
key understanding of modeling, given its extreme nature of its benefits and limitations, is to be able to
have balance, and to use models sensibly and wisely. We have to discern for ourselves which
variables are important and prudently apply them in modeling. The fact that everyone has their own
interpretation of models and parameters, the onus is on economists to explain their intuition behind
their interpretations when modeling. And in turn, it is up to readers to use their own intuition to
critique or accept the intuition. As according to Lars Peter Hansen (2014), “models are not exact
replications to reality, (and at the) end of day, it is only some type of approximation. They are
simplifications and they are not perfect. Instead of dismissing imperfect models…[he prefers] to use
them in sensible ways.” After all, models might be all that we have.
Bibliography
Bank of England, 2008, Inflation Report: February 2008,
http://www.bankofengland.co.uk/publications/Documents/inflationreport/ir08feb.pdf, last assessed 2nd
Mar 2015
Bank of England, 2008, Inflation Report: May 2008,
http://www.bankofengland.co.uk/publications/Documents/inflationreport/ir08may.pdf, last assessed
2nd Mar 2015
Bank of England, 2008, Inflation Report: August 2008,
http://www.bankofengland.co.uk/publications/Documents/inflationreport/ir08aug.pdf, last assessed
2nd Mar 2015
Bank of England, 2008, Inflation Report: November 2008,
http://www.bankofengland.co.uk/publications/Documents/inflationreport/ir08nov.pdf, last assessed
2nd Mar 2015
Bank of England, 2009, Inflation Report: February 2009,
http://www.bankofengland.co.uk/publications/Documents/inflationreport/ir09feb.pdf, last assessed 2nd
Mar 2015
Box, G. E. P and Draper, N. R., 1987, Empirical Model-Building and Response Surfaces, United
Kingdom: Wiley
Carlin, W., and Soskice, D., 2015, Macroeconomics: Institutions, Instability, and the Financial
System, United Kingdom: Oxford University Press
Hansen, L. P., 4th Dec 2014, The Consequences of Uncertainty, speech, Sheikh Zayed Lecture
Theatre, New Academic Building, London School of Economics and Political Science, United
Kingdom
15
On target: an analysis of Abenomics by Chen Yue Lok
2nd year, UCL Economics
Introduction
What can one do when the economy has been experiencing stagnation and falling prices for two
decades? Japan’s Prime Minister Shinzo Abe hopes that his response, the three-pronged Abenomics
is the correct answer that can pull Japan out of the “Lost Decades”. Japan and the world are looking
at this economic experiment with much intent, and this paper aims to answer the question on
everyone’s mind: “Is it working?” This paper attempts to do this by providing a preliminary
assessment of these policies.
Japanese economic policy is an issue worth looking into for several reasons. Firstly, Japan’s
macroeconomic environment is an unprecedentedly challenging one; stagnation for over two decades,
a deflationary period longer than that of the Great Depression of 1873 (Metzler, 2013), and a
pessimistic corporate outlook. All this is further exacerbated by a diminishing demographic dividend.
Previous policy responses have been criticized as being insufficiently aggressive in tackling Japan’s
economic woes. This is where Abenomics comes in; three massive stimulus packages, aggressive
monetary easing neither the US Federal Reserve nor the European Central Bank has dared attempt,
and structural reforms to address underlying long term issues in the Japanese economy. If Abenomics
works out, it will be credited with one of the most spectacular economic comebacks in history.
But Abenomics also has implications beyond the Japanese archipelago. Developed countries are
fearing that they themselves will experience their own lost decade in the aftermath of the recession of
2007. Policymakers in economies such as the United States and the European Union have crafted
policies which share similarities with Shinzo Abe’s package. Should Abe’s project fail, doubt will
settle on their own policy responses as well. Will Abenomics success lead to its emulation in other
countries? Or will it end up being a precautionary tale of macroeconomic folly? This is the question
this essay tries to answer.
The “Three Arrows”
i.
Monetary Policy Regime Shift
The first arrow of Abenomics brings about a shift in the monetary policy regime. A few major
changes have been done here. Firstly, Abe has instructed the Bank of Japan to set an inflation target
of 2% - a level that has rarely been hit since the early 1990s. This is a shift away from previous Bank
of Japan policy, where the Bank has vigorously resisted any calls for higher inflation (Hausman and
Wieland, 2014).
Secondly, a new bank governor has been appointed. Haruhiko Kuroda has been an aggressive critic
of the previously passive policy of the bank. He is an avid supporter of Abenomics’ aim of reflation,
and has expressed willingness to spearhead “all-out efforts to utilize every possible resource” in order
to overcome deflation (Bank of Japan, 2013).
Thirdly, the policy instrument has shifted away from the overnight policy rate to the monetary base.
The new governor has described the tools as “Quantitative and Qualitative Easing (QQE)”. The
Bank’s quantitative easing program would be the largest that has been attempted in economic history,
16
and qualitative easing in the form of expectation management aims to support the efforts to combat
deflation.
ii.
Fiscal Stimulus
The second arrow consists of massive government spending packages, in an attempt to stimulate
aggregate demand. While using fiscal policy to boost the economy is not a new idea, what sets it
apart is its sheer size. There have been three stimulus packages so far; an initial package of of ¥10.3
trillion, followed by another round worth ¥5.5 trillion in April 2014. More recently, an emergency
package worth ¥3.5 trillion was passed in December 2014, as fears of recession and deflation remerge
(The Economist, 2014). How much firepower is left in this arrow is an open question however, as the
burden of the government’s debt (around 240% of GDP) means that future fiscal stimulus will be
constrained.
iii.
Structural Reforms
Reforms to the supply side of the economy will constitute the third arrow of Abenomics. Japan’s
economic problems are closely related to rigidities in key sectors such as energy and healthcare,
relatively high costs of business and the country’s diminishing workforce due to low participation
rates from women and an aging population. As to date, progress in this area has been relatively slow,
and most of the spotlight has been on monetary policy and fiscal stimulus.
Abenomics and the IS-PC-MR model.
In order to explain the rationale and outcomes of Abenomics, this paper will use the New Keynesian
IS-PC-MR model (Carlin and Soskice, 2005). We first use the standard model to explain Japan’s
economic situation, then make some adjustments to account for the changes Abenomics has brought
about. We then use the adjusted model to check the theoretical rationale for Abenomics. For
pragmatic purposes, we will limit the modelling to a closed economy.
We start the typical case to describe Japan’s situation with a large negative demand shock (reflecting
the abrupt stop of Japan’s economic rise in 1990). The large shock leaves output below equilibrium,
and inflation below target. The central bank then forecasts the lower Phillips curve for the next
period, and attempts to set a new interest rate based on the IS curve.
Here is where conventional monetary policy in Japan is rendered ineffective. The central bank has hit
the zero lower bound on interest rates, and is unable to pick the desired interest rate to generate its
best response output gap. This suboptimal response means that output is unable to increase much, and
inflation remains low. This feeds into inflation expectations for the next period (modelled by the
further lowering of the Phillips Curve), and we can see that inflation continues to spiral downwards
for the following periods. This is the liquidity trap Japan has found itself stuck with for the previous
two decades.
In order to analyze Abenomics using the model, a few adjustments need to be made to reflect Japan’s
situation.
a. The zero lower bound will be removed from the model, giving the central bank the ability to
lower real interest rates as much as they want. This is reflected by the shift in policy
instrument from the overnight policy rate to quantitative easing, which aims to influence the
borrowing-relevant interest rate directly, rather than indirectly via the overnight rate.
b. We will start the medium-run equilibrium at a negative inflation rate, as this reflects the
“deflationary equilibrium” that Japan has been experiencing. This is also a reflection of the
Bank of Japan’s monetary policy, which has a vaguely define goal of price stability.
17
We will account for the first two arrows of Abenomics in the following ways;
a.
b.
c.
d.
The fiscal stimulus will be modelled via an upward shift in the IS curve.
The new inflation target will be reflected in a shift of the MR curve.
The usage of quantitative easing is reflected in the removal of the zero lower bound.
Supply-side policies cause an exogenous shift in the equilibrium level out output.
We start the modelling by simultaneous shifts in both the IS and MR curves, reflecting fiscal stimulus
and a new inflation target respectively. This causes current output and inflation to be lower than the
new medium run equilibrium, and the central bank forecasts the Phillips curve based on the new
medium run equilibrium and lowers interest rates in order to increase both output and inflation. As
the policy instrument is now quantitative easing, it can be said that the zero lower bound no longer
constrains the policy maker. The next period sees the economy move to a point of higher output and
inflation, and the central bank subsequently increases real interest rate gradually in following periods
until inflation is at target.
Figure 1: The IS-PC-MR model explaining Abenomics
Judging Abenomics
1. Output and Growth
Both the first and second arrows aim to revive Japan’s economic growth, and a rise in output would
definitely be welcome news. Since Abe’s introduction of his economic policies, there has been an
upward motion in both nominal and real GDP. Although the wide gap between the two underlines the
18
worries over stubbornly low inflation, Abenomics is reasonably successful in boosting Japan’s output.
This can be seen especially in nominal GDP, which is trending towards pre-crisis levels.
Figure 2: Real and Nominal GDP of Japan (Federal Reserve of St Louis, 2015)
In addition to large fiscal stimuli and quantitative easing, a few tailwinds are helping output figures as
well. A depreciating yen as well as low commodity prices may help facilitate an export-led recovery.
Private consumption and investment are also rising, albeit slowly, which is a welcome sign for the
economy. However, headwinds threaten the progress of Abenomics. A consumption tax hike in 2014
may harm prospects, although indications so far suggest that the effects will not be as devastating as a
previous hike in 1997. Secondly, a potential rebound of oil prices as a result of downwards
overshooting may negate some of the positive effects low oil prices would have brought. Global
demand still remains relatively weak, limiting the role of exports in driving the economic recovery
(International Monetary Fund, 2015).
It is difficult to mention how much of that growth is attributed to Abenomics, rather than growth that
would have happened even without the fiscal stimulus or quantitative easing. Most notably, some
have argued that the growth rates are a result of the post Fukushima disaster recovery (Hayashi,
2014). That being said, forecasting agencies including the IMF have revised their growth forecasts for
2013 upwards, citing a stronger than expected effect that Abenomics has had on Japan’s economy
(International Monetary Fund, 2013).
Other forecasters have similarly revised their forecasts
upwards, and this can be taken as an indicative sign that Abenomics can rightfully claim to have
contributed to this growth.
2. Inflation and Inflation Expectations
Deflation in Japan has not been sudden shocks, but has been rather mild, yet persistent. This has led
to arguments that Japan has now entered a new deflationary equilibrium due to secular decline in
potential growth (Nakaso, 2014). The “Conquest of Deflation” is a key aim of Abenomics, with an
inflation target of 2% to be met by 2015. The question we now ask is “Is it working?”
Looking at the data, there has been an unmistakable upward trend in inflation, and the first arrow can
claim credit in contributing to this. However, much of the inflation since 2014 has also to do with a
recent hike in consumption tax. Discount those effects, and inflation for 2014 was at a low 0.5%
according to the Bank of Japan’s own estimates. As a response to this, the bank has further expanded
its quantitative easing program in late 2014 as it races against time to meet its self-imposed deadline
of 2015 to hit their inflation target. However, falling oil prices have all but dispelled any illusion, and
19
the Bank of Japan has now revised their inflation forecast downwards, and are talking about meeting
the inflation target in 2016 instead.
Figure 3: Inflation in Japan (International Monetary Fund, 2014) The yellow line denotes the start of
Abenomics, and the green line represents estimates by the IMF.
But looking at inflation alone does not tell the whole picture; expectations of inflation are just as
important to Abenomics’ success. Our IS-PC-MR model relies on consumers adapting their
expectations (modeled by a shift in the PC curve) for the whole chain of events to work. What has
happened in Japan, however, is that inflation expectations have been stubbornly low, inhibiting the
effectiveness of previous monetary policy. A possible cause for this is the bank’s credibility has been
stained by its previous record, where policy measures to encourage inflation have been adopted halfheartedly, or rolled back prematurely (Kuttner and Posen, 2001; Ugai, 2007).
This is where
qualitative easing (as opposed to quantitative easing) by the Bank comes in, where the policymaker
acts to influence consumer expectations on inflation directly by forward guidance.
Figure 4: Inflation Expectations by Enterprises (Bank of Japan, 2014)
Looking at the Bank data from conducting surveys with companies, it does seem that the bank’s
efforts in anchoring expectations have worked. On average, Japanese firms surveyed have an outlook
20
for modest positive inflation in the short and medium run. Other measures of expected inflation such
as the inflation swap rates have also gone up after the announcement of Abenomics. This gives the
transmission mechanism for QQE much needed support (Hausman and Wieland, 2014). However,
the latest survey in December 2014 returned a marginally lower rate of expected inflation. Surely
enough, the eventually reported inflation rate was too low for the government’s comfort.
I thus infer that the Bank is unlikely to hit its self-imposed deadline for meeting 2% inflation.
However, Abe and Kuroda can take comfort in the fact that inflation expectations are relatively
favourable, and that lower oil prices may have positive effects on the economy in the medium term
despite causing deflationary pressures now.
3. Stock Market Indices
We now turn to the behavior of the Japanese stock market, namely the Nikkei index in reaction to
Abenomics. The stock market is usually seen as an indicative gauge of success of the government’s
economic policy, and we can look at the index as an expression of whether the markets are confident
in Abenomics.
Some readers may be hesitant to accept the usage of stock market indices as a reliable indicator of
economic health, as a soaring index may be camouflaging shoddy fundamentals. In particular, stock
markets primarily react to expected earnings rather than the real economy, and this may cause the
stock market to be overly volatile to be any sort of reliable indicator. However, there are a few
reasons to consider the index’s indications.
Firstly, the stock market is a relatively reliable proxy to represent investor sentiment. Should
investors feel confident in a country’s economic prospects, their reaction will reflect in a rising stock
market. By extension, this applies to confidence in a country’s economic policy as well (Cliff and
Brown, 2004).
Secondly, the stock market can influence consumption and investment via the wealth effect.
Favorable stock market conditions can increase the stockholders wealth, and consequently increase
consumption and investment (Porteba, 2000). A high stock market index may hence herald future
growth in consumption, and thus economic growth.
Nikkei 225 Index
45000.00
40000.00
35000.00
30000.00
25000.00
20000.00
15000.00
10000.00
5000.00
2015-01-04
2014-01-04
2013-01-04
2012-01-04
2011-01-04
2010-01-04
2009-01-04
2008-01-04
2007-01-04
2006-01-04
2005-01-04
2004-01-04
2003-01-04
2002-01-04
2001-01-04
2000-01-04
1999-01-04
1998-01-04
1997-01-04
1996-01-04
1995-01-04
1994-01-04
1993-01-04
1992-01-04
1991-01-04
1990-01-04
0.00
Figure 5: Nikkei Index (Federal Reserve of St Louis, 2015) The start of Abenomics is denoted by a
red line.
21
The data shows a generally upward trend of the Nikkei index since the announcement of Abenomics
in early 2013. This can be linked to Abenomics in several ways. Firstly, investors are confident in
the government’s aggressiveness to tackle economic problems, and this is reflected in bullish
sentiment in the stock market. Secondly, the depreciation of the yen due to the Bank of Japan’s
quantitative easing program is helping Japanese companies improve their export sales, thus leading to
higher profit margins. This is a good sign in the long run, as higher profits allow Japanese firms to
undertake much-needed investment that may boost the long-run capacity of the economy. Shinzo
Abe thus has good reason to feel pleased with the stock market movements as a vindication of his
namesake policy.
A note on Structural Reforms
As structural reforms are relatively long term measures, it is premature to have any comprehensive
discussion on them as of yet. However, it is clear that if Abenomics is to overcome Japan’s economic
malaise convincingly, the third arrow is arguably the most important one in Abe’s quiver.
Some structural reforms are already beginning to surface. For example, the government is attempting
to encourage more women into the workforce, thus working against a cultural background which is
disadvantageous for women who wish to start a family. Closing the gender gap can lead to substantial
economic gains, and Abe is hoping to pluck what seems to be a low-hanging fruit (Sekiguchi, 2014).
Structural reforms are also aimed at improving business conditions in Japan. Japan aims to enter the
top 3 for the ease of doing business amongst OECD countries from its current position of 15. There is
much scope for improving these conditions, such as reducing the time required to register a business,
rationalizing the registration process and reducing corporate taxes amongst other measures. It is
hoped that both domestic and foreign businesses will get a boost from these measures (Haidar and
Hoshi, 2014).
Like any reform, resistance towards Abe’s third arrow likely comes from within his own party’s
ranks. This is where Abe the politician needs to shine, to ensure that his party and the electorate are
on board with the reforms he has in mind. Any gains from the first two arrows will be rendered
meaningless if Japan’s underlying issues are not resolved.
Conclusion
To conclude, we look back at the question we first set out to answer: “Is Abenomics working?” GDP
has been on an upward trend since Abe took office, although a recent hike in consumption tax has
impeded progress. Inflation is unlikely to hit the target of 2% in the coming year, although the
government can take comfort in the fact that expectations of inflation have improved markedly. The
stock markets have also largely reacted favourably to Abenomics. The answer thus seems to be a
qualified yes, although we are arguably still in too early a stage to tell. This is especially true for the
vital third arrow of structural reforms. Nonetheless, the data so far suggests that the answer may be
“So far, so good.”
References
Brown, G. and Cliff, M. (2004). Investor sentiment and the near-term stock market. Journal of
Empirical Finance, 11(1), pp.1-27.
Carlin, W. and Soskice, D. (2005). The 3-Equation New Keynesian Model --- A Graphical
Exposition.Contributions in Macroeconomics, 5(1).
22
Haidar, J. and Hoshi, T. (n.d.). Implementing Structural Reforms in Abenomics: How to Reduce the
Cost of Doing Business in Japan. SSRN Journal.
Hausman, J. and Wieland, J. (2014). Abenomics: Preliminary Analysis and Outlook. Brookings
Papers on Economic Activity, 2014(1), pp.1-63.
Hayashi, T. (2014). Is it Abenomics or Post-Disaster Recovery?
Analysis.International Advances in Economic Research, 20(1), pp.23-31.
A
Counterfactual
International Monetary Fund, (2013). World Economic Outlook. International Monetary Fund.
International Monetary Fund, (2015). World Economic Outlook. International Monetary Fund.
Kuroda, H. (2013). Quantitative and Qualitative Monetary Easing.
Kuttner, K. and Posen, A. (2001). The Great Recession: Lessons for Macroeconomic Policy from
Japan. Brookings Papers on Economic Activity, 2001(2), pp.93-185.
Metzler, M. (2013). Introduction: Japan at an infection point. In: W. Fletcher III and P. Von Staden,
ed.,Japan's 'Lost Decade': Causes, Legacies and Issues of Transformative Change, 1st ed. Oxford:
Routledge.
Nakaso, H. (2014). The Conquest of Japanese Deflation: Interim Report.
Poterba, J. (2000). Stock Market Wealth and Consumption. Journal of Economic Perspectives, 14(2),
pp.99-118.
Pump
priming.
(2015). The
Economist.
[online]
Available
http://www.economist.com/news/finance-and-economics/21637410-shinzo-abe-unleashes-smallstimulus-package-pump-priming [Accessed 24 Feb. 2015].
at:
Sekiguchi, T. (2014). Abe Wants to Get Japan's Women Working. The Wall Street Journal. [online]
Available at: http://www.wsj.com/articles/abes-goal-for-more-women-in-japans-workforce-promptsdebate-1410446737 [Accessed 24 Feb. 2015].
Ugai, H. (2006). Effects of the Quantitative Easing Policy: A Survey of Empirical Analyses. Bank of
Japan
Working
Paper
Series.
23
The Value of Theoretical Models for Real-World Market Analysis
by Maryam Khan
Final year, UCL Economics
An analysis of whether the original Cournot and Bertrand models of oligopoly competition provide a
satisfactory basis for assessing the extent to which real-world markets are effectively competitive.
Introduction
The real-world value of economic models is a topic that is frequently disputed. The traditional
Cournot and Bertrand models of oligopoly competition attempt to explain the organization of an
economy. Yet how can their conclusions, which are based on an unrealistic world where there is
perfect information, symmetric costs, homogenous products and no strategic behavior, have any realworld significance? Many argue that models such as these should be left to textbooks as they have
little value in real-world analysis. Models are designed to explain complex observed processes and are
therefore subjective approximations of reality. They can never be perfect but using them in economics
is imperative even though it has led to the “assume we have a can opener” catchphrase1 to mock
economists and other professionals who base their conclusions on unrealistic or unlikely assumptions.
On the one hand, if we are to disregard theories because they rely on unrealistic assumptions then
economics (and other professions) won’t be left with much. Furthermore, what defines an unrealistic
assumption from a realistic assumption? On the other hand, the very process of constructing models,
and then testing and revising them forces economists and policymakers to tighten their views about
how an economy works. After all, Milton Friedman argued that theories with accurate predictions are
of great value even though their assumptions may be extremely “unrealistic.” This paper analyses the
traditional Cournot and Bertrand models in order to assess their direct applicability in market analysis,
therefore determining their subsequent value. It is important to note that “value” in itself is a
subjective term and is by no means limited to how well a model can be applied to the real world.
Whilst the Cournot and Bertrand models may have limited real-world applicability, their value arises
from the knowledge they have given economists and regulators about industrial organization and
market power in an oligopoly.
An oligopoly market lies in the middle of the extremes of a monopoly and perfect competition. We
know that the outcome of perfect competition is price equal to marginal cost and the outcome of a
monopoly is marginal revenue equal to marginal cost. The main difference of the oligopoly market is
that firms are assumed to take into account interdependencies, which means that their decisions have
taken into account what their rivals might do. Moreover we assume that there are many consumers
(none large relative to the size of the market), few firms and barriers to entry. At the most basic level
(the traditional models of Cournot and Bertrand), decisions are made simultaneously and the product
is assumed to be homogeneous. The outcome of an oligopoly market depends upon the strategy,
which in turn depends upon the model we are looking at or “the game” that firms are playing when
they are competing. Cournot and Bertrand models of oligopoly competition are examples of two
1
The story is as follows: A physicist, chemist, and an economist are stranded on a desert island with no implements and a can of food. The
physicist and the chemist each devised an ingenious mechanism for getting the can open; the economist merely said, "Assume we have a can
opener"! (Boulding, 1970)
24
different games firms could be playing as the former focuses on quantity competition and the latter
price competition.
Competition authorities recognise the limitations surrounding these models but use their predictions
as a basis for theories of harm in competition investigations. Specifically competition authorities rely
on oligopoly models to look at the impact of competition between a few firms when there are barriers
to entry. For example regulators may map the structure of the market under question to the model
assumptions to identify which model is the “best fit”. From this, they consider under which
circumstances the model would predict a welfare concern. They are also used in merger analysis, to
see whether a proposed merger poses a competition concern. Moreover the traditional Cournot and
Bertrand models provide a useful “rule of thumb” and help to set a benchmark for analysis, explaining
what can happen in different types of markets.
Once we can establish what happens in the simple Cournot and Bertrand world, we can move on and
tweak the assumptions to develop more complex models, which then help to get a closer perspective
of reality. Hence, although Cournot and Bertrand themselves may not be very good at explaining the
real world, they should not be ignored when it comes to learning about oligopoly competition as they
provide a foundation for more in-depth analysis.
The Cournot model
The Cournot model or Cournot duopoly is named after Antoine Augustin Cournot (1801-1877) who
was inspired by observing competition in a spring water duopoly (Varian, 2006). It refers to the game
whereby there are two firms producing a homogenous product. Each firm faces identical/symmetric
costs and there is no co-operation and no entry. These two firms compete simultaneously by choosing
output. As they take interdependencies into account, these two firms have reaction functions whereby
they make optimal, profit-maximising quantity decisions based on what the other firm is doing. A
point on the firm’s reaction function is the best response for that firm, given what the other firm is
doing. As each firm has a reaction function, the outcome of this model is a Nash equilibrium where
each player in the game has selected the best response (or one of the best responses) with regard to the
other players' strategies (Nash, 1950).
Algebraically, we can solve the general case to show the outcome of a Cournot model. If we begin by
supposing there are N firms. Price is the same across these N firms and is determined by market
demand (Q), where Q =∑𝑖=1…𝑁 π‘žπ‘– .
P = a-bQ = a-b∑𝑖=1…𝑁 π‘žπ‘– where a and b are constants.
We assume symmetric costs for both firms: C(π‘žπ‘– ) = cπ‘žπ‘– .
From this we can work out total revenue and individual firm profit by incorporating the residual
demand, which is the market demand that is not met by other firms in the industry. In particular,
quantities are “strategic substitutes”; if one firm increases output, another firm has lower residual
demand for that price.
πœ‹i = TR-TC = Pπ‘žπ‘– − cπ‘žπ‘– = (a-b∑𝑖=1…𝑁 π‘žπ‘– )π‘žπ‘– − cπ‘žπ‘– = (a-c) π‘žπ‘– - bπ‘žπ‘– 2 - bπ‘žπ‘– ∑𝑖≠𝑗 π‘žπ‘—
As firms are said to be profit maximising, we can differentiate the equation above and set it equal to
zero:
πœ•π‘žπ‘—
πœ•πœ‹i
= (π‘Ž − 𝑐) − 2π‘π‘žπ‘– − 𝑏 ∑ π‘žπ‘— − π‘π‘žπ‘– ∑ (
)
πœ•π‘žπ‘–
πœ•π‘žπ‘–
𝑗≠𝑖
25
𝑗≠𝑖
We assume that
πœ•π‘žπ‘—
πœ•π‘žπ‘–
= 0 by the zero conjectural variation, which says that if a firm (for example firm
A) makes an output decision, we can hold output choices of other firm’s (firm B) constant. In other
words, firm B won’t change it’s output choice upon hearing A’s output choice.
Therefore :
πœ•πœ‹i
= (π‘Ž − 𝑐) − 2π‘π‘žπ‘– − 𝑏 ∑ π‘žπ‘— = 0
πœ•π‘žπ‘–
𝑗≠𝑖
This rearranges to give:
2π‘π‘žπ‘– = (π‘Ž − 𝑐) − 𝑏 ∑ π‘žπ‘—
𝑗≠𝑖
π‘žπ‘– =
(π‘Ž − 𝑐) − 𝑏 ∑𝑗≠𝑖 π‘žπ‘—
2𝑏
This is firm i’s reaction function. As we can see it depends upon firm j’s output choice.
If we assume there are two firms (A and B) we can draw the reaction functions of each and the
intersection is where the equilibrium outcome is.
Moreover if we assume symmetry, then each firm produces the same:
q1=q2=….qN
Then we get:
2bπ‘žπ‘– = (π‘Ž − 𝑐) − 𝑏(𝑁 − 1)π‘žπ‘–
π‘žπ‘– =
(π‘Ž−𝑐)
𝑏(𝑁+1)
This is the outcome decision for individual firms.
As Q is the total output of the economy and we have assumed symmetry, therefore 𝑄 = π‘π‘žπ‘– :
𝑄 = π‘π‘žπ‘– =
𝑁(π‘Ž − 𝑐)
𝑏(𝑁 + 1)
This is aggregate output of the oligopoly market.
To find price we can substitute this into P = a-bQ to get:
26
𝑃 = (π‘Ž − 𝑏𝑄) = π‘Ž − [
𝑏𝑁(π‘Ž − 𝑐)
𝑁(π‘Ž − 𝑐)
π‘Ž + 𝑁𝑐
]= π‘Ž−[
]=
(𝑁 + 1)
𝑏(𝑁 + 1)
𝑁+1
We arrive at the same outcome if we were to simplify P=a-bQ to P = a-Q (i.e. b=1):
𝑃 = (π‘Ž − 𝑄) = π‘Ž −
𝑁(π‘Ž − 𝑐)
π‘Žπ‘(𝑏 − 1) + π‘Žπ‘ + 𝑁𝑐
=
𝑏(𝑁 + 1)
𝑏(𝑁 + 1)
If b = 1 this simplifies further to P =
π‘Ž+𝑁𝑐
𝑁+1
The result of this model is therefore one equilibrium point where firms have no incentive to deviate
from; each firm produces the same quantity and charges the same price. It can be shown that the
Cournot outcome lies in between the perfect competition and monopoly outcomes. This is useful for
competition authorities as it allows them to compare this case to perfect competition and monopoly
outcomes and to see the effect on welfare (by calculating the deadweight loss). By changing the
number of firms, the effect on welfare and market power can be scrutinised. For instance, further
analysis highlights how increasing the number of firms (N), causes the market to perfect competition.
This can be useful for competition authorities when they are assessing the effect of, say, a merger.
The main limitation of this model is the naiive conjectural variation assumption which states that
πœ•π‘žπ‘—
πœ•π‘žπ‘–
= 0, or in other words that firm i won’t change it’s output choice upon hearing firm j’s output choice.
Realistically, a rival may very well change output upon learning about the output decision of another
firm. In addition, in the real world, firms do not face symmetric marginal costs and they do not
compete simultaneously. There may be certain markets where this may arise, but in general this isn’t
the case for most marekts. The Cournot model also ignores the fact that firms might co-operate and
collude with each other to reach a more profit maximising outcome. Therefore the Nash equilibrium
identified above, won’t be true when applied to the real world.
However, out of these concerns, more complex models have been born, such as the Stackelberg game
which features sequential moves (a leader and a follower). The outcome of this game is that the leader
produces more than the Cournot equilibrium and the follower produces less than the Cournot
equilibrium. In addition, we have looked at N firms, an even more basic Cournot model only
considers two firms. Therefore there is vast scope for advancing the Cournot model (as mentioned
later) to better explain the real world.
The Bertrand model
The Bertrand model, named after Joseph Louis Francois Bertrand (1822-1900), describes a game
where firms set prices and it is the consumers that choose quantities at the prices set. Bertrand
formulated this model in a review of the Cournot model and found that when firms set prices, the
optimal outcome is similar to that of a perfectly competitive world (price equal to marginal cost).
Once again this model relies on strict assumptions: homogeneous products; no collusion;
simultaneous decision-making and symmetric costs. Similar to the Cournot conjecture, the Bertrand
conjecture states that in equilibrium, other firms won’t want to change their price choice in response
to Firm i’s price decision (Edgeworth, 1925).
In order to see the outcome of this game, we assume there are two firms (firm A and firm B).
Although we know that they move simultaneously, in order to arrive at a price decision, each firm
thinks sequentially. For instance, firm A thinks about making a profit-maximising price decision,
taking into account what firm B will respond with. If firm A sets a high price (above marginal cost),
firm B will react by charging a slightly lower price (undercutting) and capturing all of the market. As
27
goods are homogeneous, consumers buy from the lowest cost firm. Thus, firm A will then react by
undercutting B and this will continue until both arrive at price equal to marginal cost. As they have
set the same price, demand is split evenly between them and they each capture half of the market.
Therefore this Bertrand-Nash equilibrium is where Pa=Pb=MC and qa=qb=Q/2. The logic is that if A
or B charged a price above marginal cost, no one will buy from them and neither would set a price
below marginal cost as it would incur a loss (they would rather shut-down and leave the market).
This model is useful because price competition is observed more often than quantity competition.
However, this outcome isn’t very applicable to a real-world market, as it is based upon unrealistic
assumptions that don’t arise in most markets. Namely, the zero conjectural variation as seen with the
Cournot model, which states that firms won’t change their price upon hearing their rival’s price.
There are many pricing strategies firms may employ that might result in the Bertrand conjecture not
holding, such as predatory pricing (where firms deliberately price below their marginal cost, thereby
incurring a loss in order to drive rivals out of the market). Furthermore, this model also relies on the
assumption that the consumer will always buy from the cheapest firm. Behavioural economists can
name a wide variety of reasons as to why this may not be the case; for instance, the quality may be
perceived to be greater with a higher price. Another reason could be that consumers are not fully
informed and may not necessarily know that there are cheaper options elsewhere.
Do these models hold any value?
Although these models may not be very effective at describing the real world, they do provide a
useful benchmark for competition authorities to assess the market with. These traditional oligopoly
models, when compared to perfect competition and monopoly cases allow economists to establish a
“rule of thumb” which aids competition investigations. For instance, more firms are better (as we get
closer to perfect competition) and price competition is a good thing (seen by Bertrand model giving
the perfectly competitive outcome of price equal to marginal cost).
The assumptions of perfect information, simultaneous moves and homogenous products limit the
applicability of both Cournot and Bertrand models in real-world market analysis. These traditional
models have been extended to include asymmetric information as well as sequential moves and
product differentiation. Each underlying assumption of the traditional oligopoly models can be
changed and the resultant model gets closer to the real world. For instance, Kreps and Scheinkman
(1983) brought the two models together to provide a middle ground whereby firms can compete on
both quantity and price by choosing a capacity constraint in the first period and a price in the second
period. The result is, surprisingly, a Cournot outcome. Hence in this case, it is very important to learn
about the Cournot model, as more complex cases may revert back to the Cournot outcome as seen
above. Additionally, Cournot and Bertrand can be modified to deal with heterogeneous costs or
exogenous product differentiation. The Stackelberg game deals with sequential moves and models
also deal with endogenous product differentiation (such as representative consumer models of
monopolistic competition and locational models). These more advanced models deal with the
limitations of the traditional Cournot and Bertrand models by removing one main assumption and
analysing the effect on the outcome. They get closer at explaining the real world, but as they still
involve assumptions, their direct applicability is still somewhat limiting. Whether this makes them
more valuable and useful than the traditional Cournot and Bertrand models is limited by our definition
of “value”. If we are referring to how well these models describe the real world, then sceptics may be
justified in their criticisms; perhaps they should be resigned to textbooks. However if value
incorporates the usefulness to market analysts, the predictive nature of these models, as well as how
they’ve been enhanced to explain more complex economic phenomenon, then their stringent
assumptions are their only drawback.
28
Competition authorities face a trade off when analysing the market, as they want to carry out in-depth
analysis in order to arrive at justifiable recommendations but have limited time and resources.
Including private information about costs can make the analysis very cumbersome for competition
authorities. For the purpose of establishing theories of harm, the traditional Cournot and Bertrand
models overcome the trade off competition authorities face with having limited time and resources to
analyse the market and arrive at justifiable recommendations.
Although the Cournot model doesn’t necessarily give a realistic view of the world due to its limiting
assumptions, it does give competition authorities an idea of whether there might be a competition
concern, by allowing them to compare outcomes to a monopoly and perfect competition. Therefore it
helps them form theories of harm when beginning the investigation procedure, allowing them to
concentrate their analysis on these specific concerns. This makes the investigation process more
efficient and less time-consuming as competition authorities are only investigating cases where there
is an initial concern.
Conclusion
To conclude, I agree that Cournot and Bertrand offer a biased explanation of the real world, as each
are constrained by rigid assumptions which limit their application to reality. However, learning about
them is vital as they allow us to develop more complex models, and more importantly they aid
competition authorities in assessing whether there are competition concerns that need to be
investigated. Although neither traditional models of Cournot or Bertrand have significant direct
practical applicability to the majority of real-world markets, these theorems and the outcome they
arrive at do have real-world market value as they provide a useful benchmark from which to look at
the market. Competition authorities use them to establish theories of harm when carrying out
investigations. It is highly unlikely that we’ll ever arrive at a model that will fully explain and help us
manipulate our complex world. However the value of these traditional economic models and theories,
as Friedman set out in his ‘Methodology of Positive Economics’, lies in the accuracy of their
predictions. He maintains that the realism of a theory's assumptions is irrelevant to its predictive
value. It does not matter whether the assumptions that firms maximize profits or move simultaneously
are realistic. Theories and models should be appraised exclusively in terms of the accuracy of their
predictions. The accuracy of each model depends on how well it imitates the industry in question;
Bertrand will be better if capacity and output can be easily changed (firms are competing on price)
whilst Cournot is generally better if output and capacity are difficult to adjust (and firms are
competing on quantity). Thus their value comes from the fact that they form the foundations of our
knowledge into how firms behave in oligopoly markets. Moreover it is from the traditional Cournot
and Bertrand models that more complex models have been developed; therefore in order to learn and
understand the complicated real world, one must be able to understand what happens in a simplified
world.
References
Boulding, K. E. (1970). Economics as a Science.
Edgeworth, F. (1925). The pure theory of monopoly. Collected Papers relating to Political Economy.
Nash, J. (1950). Equilibrium points in n-person games. Retrieved 2014, from Proceedings of the
National Academy of Sciences of the United States of America:
Varian, H. (2006). Intermediate microeconomics: a modern approach (7 ed.). W. W. Norton &
Company
29
A Dynamic, Hierarchical, Bayesian approach to Forecasting the 2014
US Senate Elections by Roberto Cerina2
3rd year, UCL Economics & Statistics
Introduction
This paper sets out to accomplish 2 objectives:
1.
2.
to accurately predict the 2014 Senate election-day Republican vote share;
to grasp the lessons from the earliness/accuracy trade-off.
I will show that the 2014 Senate election was uncertain up to the very end, and Republicans’ fortunes
turned in the last few weeks. A bias in the polls favoured the Democratic party and led to unexpected
uncertainty over the results. This uncertainty was gone by the final week of polling and a Republican
takeover of the Senate was close to unquestionable. The extent to which the Republican wave
demonstrated itself was, however, unexpected.
In 2012, Drew A. Linzer presented a dynamic hierarchical Bayesian state level forecasting model to
predict the US Presidential election. The model performed remarkably in predicting the election
results, but also had the unique feature of providing estimates for voter behaviour throughout the
campaign.
Here I present an adaptation of Linzer’s model. My model’s peculiarity is that it focusses on state
level structural fundamentals for its historical forecast, as opposed to national factors, and adopts a
full-Bayesian specification to propagate the uncertainty that exists on these structural variables onto
my final predictions. The model combines historical trends with increasingly precise polls.
At the end of every week during the election campaign, I take a summary of published polls and fit
the model to calculate the distribution of voter preferences for every week in the campaign, up to
election-day. Initially, the predictions are as good as the historical forecast, but they get closer to the
actual election results as more and more polls are encoded. Simulation techniques are then used to
calculate the probability that Republicans take over the Senate.
Methodology
The Bayesian approach to statistics is philosophically different to the more main-stream “frequentist”
paradigm. A full Bayesian model allows me to integrate multiple datasets in a single model; estimate
regression coefficients in a Bayesian fashion, enabling me to encode prior knowledge in each; and
propagate the corresponding uncertainty all the way to the posterior distribution of the Republican
vote share.
Frequentist approaches to prediction using polls are available, however they pose problems in
interpretation (the p-value doesn’t quite match a Bayesian probability) and unsatisfactory in
dynamics, since they do not easily allow us to estimate voter behaviour on every day of the campaign.
For the remainder of this section, I will outline the full Bayesian model. I make extensive use of the
2012 paper from Drew Linzer [14], which served as inspiration for this piece of work.
2
Project Supervisor: Dr. Gianluca Baio
30
1. Distribution of the “trial heat" polls
My first task is to model the trial heat polls. Let 𝑖 be the state index defined over: 𝑖 = 1, . . . ,33 and let
𝑗 be the time index, representing the last 22 weeks of the election campaign: 𝑗 = 1, . . . ,22. Let 𝑛𝑖,𝑑,π‘˜
be the sample size of trial heat poll π‘˜, in state 𝑖, on day 𝑑 of the campaign; let 𝑦𝑖,𝑑,π‘˜ be the number of
respondents who declared they would vote for the Republican candidate, in state 𝑖, for poll π‘˜, on day 𝑑
of the campaign. Then π‘Œπ‘–,𝑗 is the sum of preferences 𝑦 over all polls π‘˜ in week 𝑗 and 𝑁𝑖,𝑗 is the sum of
sample sizes, over that same week: 𝑁𝑖,𝑗 = ∑π‘˜ 𝑛𝑖,𝑑,π‘˜ and π‘Œπ‘–,𝑗 = ∑π‘˜ 𝑦𝑖,𝑑,π‘˜ . This allows us to present our
distribution for the data:
π‘Œπ‘–,𝑗 ∼ Bin(𝑝𝑖,𝑗 , 𝑁𝑖,𝑗 ).
(1)
The parameter 𝑝𝑖,𝑗 is the one we are trying to estimate and it represents the probability that a citizen to
vote for the incumbent senator, for any state, at any given week of the campaign.
2. Prior distributions
I define the Republican vote share as follows:
𝑝𝑖,𝑗 = logit −1 (πœπ‘–,𝑗 + πœ›π‘— ).
(2)
I adopt a hierarchical specification to explain the distribution of the vote share, where πœπ‘–,𝑗 represents
the state specific effect on any day of the campaign, whilst πœ›π‘— represents the common trend amongst
Republican candidates at the national level. I call this the Republican campaign effect.
I am confident that as election day nears, the candidates’ campaign efforts will become superfluous as
most voters will have already made their mind up by that time[11]. Because of this, I feel it is
appropriate to assume that the Republicanvote share will not be affected by national campaign effects
on election week. I therefore fix
πœ›π½ ≡ 0,
(3)
where 𝐽 is election week (the last week of the campaign). I know that voters behaviour during the
campaign is temporally dependent, i.e. we would expect an individual’s preferences in week 𝑗 + 1 to
be dependent on his preferences in week 𝑗 . This suggests that a random walk process would
adequately specify the temporal distribution of the campaign effect. This allows me to bridge the gap
between the last released polls and election week, when estimating the election week forecast weeks
or months in advance [21]. Since we fixed the election week campaign effect at 0, we make the
random walk begin there and walk backwards:
2
πœ›π‘— |πœ›π‘—+1 ∼ 𝑁(πœ›π‘—+1 , πœŽπœ›
).
(4)
I also believe that the impact of the campaign could be of any magnitude during the election and my
model should not, a priori, be given a small variance which would assign higher probability density to
values around πœ›π‘—+1 . As a result, it makessense to give a non informative prior to the standard
deviation of the random walk:
πœŽπœ› ∼ Uniform(0,10).
(5)
My belief for the distribution of the state level effect is rather more complicated. Because of the
literature available on “economic voting", I can get a good estimate for the state level Senate popular
vote, on election day, through an adaptation of the TFC model. My version of this model is based on
state level structural economic factors, π‘₯1 ; a state level dummy variable representing the incumbency
of a candidate, π‘₯2 ; the incumbent president’s approval rating, π‘₯3 ; a dummy variable representing
31
affinity of the incumbent party candidate with the incumbent president, π‘₯4 . I therefore regress the
incumbent party vote share β„Ž on the variables mentioned above, for every state 𝑖 = 1, . . . ,33. I adopt a
logistic specification sincethe incumbent party faces two possibilities: re-election or defeat. The
regression looks like this:
logit(β„Žπ‘– ) = 𝛼 + 𝛽1 π‘₯1,𝑖 + 𝛽1 π‘₯2,𝑖 + 𝛽1 π‘₯3,𝑖 + 𝛽1 π‘₯4,𝑖 + πœ–π‘– .
(6)
Instead of performing a standard regression with the usual logistic distributional assumptions, I decide
to perform the regression in a Bayesian fashion. Because of this, I need to express my prior belief
upon the distribution of the coefficients π‘‘π‘Ž, the general error term 𝛼 and the state specific error term πœ–.
Here I encode vague priors on all the regression and let the data speak for themselves:
𝑖𝑖𝑑
𝛼, 𝛽, πœ– ∼ N ormal(0, 𝑣).
(7)
The advantage of using a Bayesian specification, even without meaningfully informative priors, will
become clear in the estimation phase: the uncertainty carried by these parameters will propagate to
our posterior Republican vote share probabilities.
It follows that the state level effect should be fixed to my historical estimates on election week,
normally distributed with precision parameter 𝜏 representing my degree of belief in the TFC model,
where πœπ‘– =
1
π‘ β„Ž2
:
𝑖
πœπ‘–,𝐽 ∼ 𝑁(logit(β„Žπ‘– ), π‘ β„Ž2𝑖 ).
(8)
In order to specify an appropriate value of π‘ β„Ž2𝑖 , I use a sensitivity analysis made by Linzer [?], which
suggests a precision level 𝜏 not exceeding 20.
Similarly to the national effect, I want to learn how voter behaviour as related to the structural
forecast evolved during the campaign. I apply the same logic as above and assign a reverse random
walk prior to the state level forecast:
𝑣_𝑗|πœπ‘—+1 ∼ Normal(πœπ‘—+1 , 𝜎𝜐2 )
(9)
where the estimated variance 𝜎𝜐2 represents the rate of weekly change in state effect. 𝜎𝜐2 is assigned a
non-informative prior, because I cannot know a priori what the behaviour of the weekly rate of
change in the state level effect is going to be. This rate is dependent on a chain of random events
including a successful party convention; performance in the debates; gaffes; endorsements etc. such
that the uncertainty about its magnitude is best expressed with a uniform prior:
𝜎𝜐 ∼ Uniform(0,10).
(10)
3. Estimation
We update the historical forecast in a Bayesian fashion, for any point in time during the campaign.
We use Bayes theorem:
Pr ( 𝑝|π‘Œ) =
π‘ƒπ‘Ÿ ( π‘Œ|𝑝) π‘ƒπ‘Ÿ ( 𝑝)
.
π‘ƒπ‘Ÿ ( π‘Œ)
(11)
The problem we face here is that the above calculation is rather cumbersome, because of the nonconjugated nature or the prior distributions. Thankfully, there are far more efficient ways of
calculating the posterior.
32
“Monte Carlo Markov Chain" (MCMC) methods are available to us through a series of algorithms,
whose purpose is to aid us in finding the posterior distribution of a population parameter of interest. I
use the JAGS [19] software through the emphR2jags package for R [10] to access the algorithms.
4. Interpretation
My forecast for time 𝐽 will be some sort of compromise between the most recent polls and
the historical forecast. If we are trying to fit the model close to the election week, πœ›π‘— will be
already close to 0, πœπ‘–,𝑗 will not have much time to revert back to the historical forecast so 𝑝𝑖,𝑗
will be determined largely by the polls. Instead, if we are trying to fit the model months away
from the election day, the opposite is true, with the 𝑝𝑖,𝑗 being determined largely by the
structural forecast (as we lack polls for most of the campaign days). The precision parameter
𝜏 determines the speed of convergence towards the historical forecast.
Because of the random walk specification, older polls are discounted in weight, but they
leave behind their historical estimates of 𝜐 and πœ› , throughout the campaign. This is
interesting as it allows us to look at how voter behaviour evolved during the campaign. The
historical set of estimates for πœ› shows the magnitude and direction of national campaign
effects. By comparing πœπ‘–,𝑗 with πœ›π‘— we can have a first hand look at how much of the changes
in voter behaviourwas due to national campaign factors.
Application:2014 US Senate
In this section the model’s predictions are presented and compared to actual election outcomes. An
evaluation will follow, based on different measures of comparison.
1. Campaign unfolding
In Figure 1 we can see an overview of the model’s performance. Out of the 33 races, only North
Carolina proved to be on the wrong side of the 50% vote share line, suggesting that the model does a
pretty good job at predicting the “sign” of the election.
The predicted probability of a Republican senate takeover, according to the model, was 94% by the
end of election week. The most probable outcome is predicted to be a Republican win of 19 seats —
or a net gain of 7 seat, 1 more than they need to take over the Senate. Figure 2 depicts the
overwhelming election day Republican advantage.
33
Figure 1. We compare the predicted two-party vote share distribution with the actual results (and the
state-level TFC predictions). The prediction interval is reported as 2 s.d. around the predicted mean
Republican vote share. A safe Republican seat is coloured red and is defined as the mean being at least 2
s.d. greater than the 0.5 cut-off; a likely Republican seat is coloured light red and is defined when the
mean is larger than the 0.5 cut-off, but the lower tail touches the line. Democratic seats are defined in the
same way, with a blue colour scale. The colouring of the labels on the y axis indicates the incumbent
party: Republican if red and Democratic if blue.
34
Figure 2. A histogram showing the probability density of the number of seats the Republican party is
likely to win in the 2014 Senate election. The number of seats represented by each column is displayed on
the x-axis, under the rightmost corner of the column.
The model suggests that whilst the election was initially to be perceived as close, but more probably
Democratic, it became a toss up as the Republican campaign successfully turned voters around. The
last week of polling represented the coronation to a successful campaign, with key races becoming
more certainly Republican. In Figure 3 one can see the evolution of the probability of a Republican
win.
Figure 3. The evolution of the probability density assigned to the number of seats to be won by the
Republicans. Starting from the top-left, we see a snapshot of the probabilities when there are 6 weeks left,
then 4 and so on, up to election-day.
35
To show how the model can be used to illustrate the unfolding of the election, I will analyse North
Carolina and try to explain changes in voter behaviour with substantial historical evidence from
campaign events. Analysing what happened here, one can immediately see that the structural model
for North Carolina is solid: it reduces our uncertainty of the entire race to less than 10 percentage
points, with the Republican at a Slight disadvantage, all the way up to week 7 (15 weeks to go).
An odd ball in this race, was represented by Libertarian candidate Sean Haugh, who polled
vertiginous high for a third party candidate all the way up to election day[22]. His presence breaks the
assumption that allowed us to model the 2-party vote share, which is that it is ok to only consider
Democrats and Republicans, as long as the assumption that independents, third party candidates and
non-respondents in polls break evenly for both parties.
Tillis, the Republican contender, managed to put on a good show in the last tv debate (around 2 weeks
to election day), as Hagan was a no-show in the televised debate amid criticism over her husband
reaping personal benefits from the Obama stimulus package [16]. This propelled him closer to Hagan,
but he never quite caught up in the polls. On election day, he won with a 1.5% lead.
At the national level, it is interesting to see how the Republican national campaign impacted the race.
It seems that initially, the name of the Republican party was rather unpopular, and on average started
at a disadvantage. However the campaign seems tohave progressed smoothly for the “Grand Old
Party", as they manage to erase this national disadvantage and eventually come out on top of the
contest.
Finally, full convergence is achieved every time the model is fit, and the Gelman-Rubin statistic is
always consistently under the 1.1 mark. As we try to fit the model earlier in the campaign, the model
fatigues to converge a bit more, and requires more computational time.
36
Figure 4. Explaining voter behaviour in North Carolina. This graph shows weekly changes in the
prediction interval of the Republican vote share, at three different stages of the campaign. The incumbent
Democrat Kay Hagan was unexpectedly defeated by Republican challenger Thom Tillis
37
Figure 5. The Republican Campaign effect. The drop in week 15 coincides with the low approval ratings
of the Republican controlled congress, which impacted the campaign and polls. [9]
Figure 6. Convergence in the model: all parameters sit comfortably under the 1.1 mark of the GelmanRubin statistics, indicating convergence.
38
2. Bias in the polls
The real outcome of the election was in fact a net gain of 8 seats, or a gross total of 20 seats, which
was rather improbable under my model. An investigation into reasons for such an under-estimation
reveals that polls were heavily skewed in favour of the democrats. In their own evaluation, Yougov
calculates the average Democratic bias from some top pollsters, and it turns out the bias was around
2.8%[17]. This is shown in Table 1.
Pollster
Suffolk
Monmouth
University
Survey USA
CNN
Public Policy
Polling
YouGov
NBC/Marist
Rasmussen
FOX
Democratic Vote
Number Percent Average Mean
of Polls Correct Error Absolute
Error
6
7
83%
71%
0.11
2.33
2.03
2.45
10
7
12
80%
71%
83%
2.56
1.90
2.77
2.66
2.71
2.91
35
9
12
7
94%
78%
75%
71%
1.99
3.36
2.54
2.69
3.18
3.36
3.37
3.58
Table 1. Average bias in favour of the Democratic party, from 9 top pollsters, as calculated by YouGov
[24]
Year
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
2014
Partisan bias in the
Senate polls
Average Favoured
bias
party
1.0
1.1
3.1
0.2
4.9
2.1
4
0.2
2.7
0.2
0.9
3.4
4.0
D
D
D
R
R
R
D
D
R
D
R
R
D
Table 2. Depiction of the average statistical bias in polls conducted during the last 21 days of the election
campaign, from 1990 to 2014, as calculated by fivethirtyeight.com [5].
39
Was there a conspiracy to favour the Democrats in polls? Not quite. Polls bias, as shown by Nate
Silver on fivethirtyeight.com [23], is actually rather common, and random on average. To argue this
point, he calculates the average polls bias in the last elections since 1990, as shown in Table 2.
Sam Wang, at the Princeton Election Consortium, points us to an interesting explanation for the bias:
polls were biased because democrats didn’t turn up for the mid-term election. Figure 7 shows a
correlation between underestimation of the Republican voteand low turnout. This thinking seems to
be in line with political pundits, who argue that what they call the “Obama coalition” does not turn up
when the current president is not on the ballot [18].
[24]
Figure 7. The plot shows an apparent correlation between low turnout and underestimation of GOP
performance. Did the Democrats stay at home? And why did the pollsters fail to account for that?
3. Evaluation
In this subsection I will refer back to the 2 objectives that I set out to achieve in the introduction, and
try to evaluate the performance of the my forecast in meeting those.
Starting with (1): Did I manage to accurately predict the election day 2-party vote share? The first
measure I will use to evaluate the accuracy of my election day forecast is the Brier Score. The Brier
score is a measure of accuracy for predictions of mutually exclusive, categorical outcomes [6]. Table
3 gives us a ranking of predictors, as calculated by Sam Wang and by The Washington Post.
Brier Score Calculations
Washington Post
Sam Wang (Princeton Election
Consortium)
Daily Kos
Washington Post
FiveThirtyEight
PredictWise
Huff. Post
Upshot
Princeton
0.024
0.027
0.032
0.032
0.034
0.035
0.043
Daily Kos
Washington Post
FiveThirtyEight
Betfair
Huff. Post
Upshot
Princeton
0.10
0.12
0.14
0.14
0.14
0.15
0.18
Table 3. Ranking of pollsters by Brier Score, a measure of prediction accuracy. My model would be third
in the rankings as calculated by Sam Wang, worse than the Washington Post but better than
fivethirtyeight.com. Drew Linzer, at the Daily Kos, takes gold again [13].
40
With a Brier Score of 0.031, I sit comfortably top 3. This low score is in line with what we saw in
Figure 1, where the direction of the red wave was well captured by the model.
Another choice of score statistic is the Root Mean Squared Error (RMSE), which tells us on average,
by how many points we missed the true values of the vote share. This suggests that, on average, I miss
the correct percentage vote share by 2.8%. This result is not ideal; however, it should be put into the
context of the skewed polls scenario.
(2): What can we learn about the election from the trade off in terms of accuracy and earliness?
The Brier Score inflates ever so quickly, levitating every week, from 0.031 to eventually 0.071, when
the model is fit 1 month before the election. The average weekly growth of the Brier Score is 56%.
The RMSE follows a similar pattern, however itseems to be ever so slightly more robust, reaching
0.038 1 month out, and demonstrating a weekly average growth rate of 26%.
A month away from the election, just under 50% of the races carried uncertainty in terms of which
party is going to win. Just 2 weeks later, when the probability of a Republican win is still only 45%,
the uncertain races are exactly 6. The week after that, the probability of a Republican win becomes
close to 94%.
This scenario suggests an election which was decided by a few close races that changed front-runner
during the last month of the election. These close races contribute to heavily move the brier score, as
even if the prediction of the percentage vote stays close to the actual result, it changes the side of the
50% line on which it falls. Furthermore, the model isn’t actually able to grasp the full weight of the
swing, falling short of predicting the swing in North Carolina and the extent to which the other key
states had changed their minds. This is an important conclusion, as it suggests that there is an element
of last minute unpredictability in the behaviour of voters.
Conclusion
The dynamic Bayesian approach to election forecasting has proved to be a reliable tool in this thesis,
succeeding in producing trustworthy predictions even for an uncertain election such as the 2014 race
to the Senate.
A topic for further research would be to expand the hierarchical specification to also include regional
effects, such as a “South", “Great Lakes" or a “New England" effect on the Republican vote. Other
tweaks can be made to the distribution of the polling data, and to the structurathel forecast, to include
third party candidates and independents. Further research should also look into how we can have state
specific priors for determining the changes in voter behaviours.
Elections are not random phenomena, but rather an expression of specific governing dynamics, which
we can understand, explain and, eventually, engineer. Whilst this is a frightening prospect for the
future democracy, it promises to be an inexhaustible source of inspiration for the next generation of
statisticians.
References
[1] Alan Abramowitz. Forecasting in a polarized era: The time for change model and the 2012
presidential election. PS: Political Science & Politics, 45(04):618{619, 2012.
[2] John Adams and Charles Francis Adams. The Works of John Adams, Second President of the
United States: With a Life of the Author, Notes and Illustrations, volume 2. Little, Brown, 1850.
41
[3] Christopher J Anderson. Economic voting and political context: a comparative perspective.
Electoral Studies, 19(2):151{170, 2000.
[4] Gianluca Baio. Bayesian methods in health economics. CRC Press, 2012.
[5] Carl Bialik. The polls were skewed toward democrats. FiveThirtyEight.com, 2014.
[6] Glenn Wilson Brier. Veri_cation of a forecaster's con_dence and the use of probability statements
in weather forecasting. US Department of Commerce, Weather Bureau, 1944.
[7] Encyclopedia Britannica. Two-party systems. Online, April 2014. URL
-party/36668/Two-party[8] William F Christensen and Lindsay W Florence. Predicting presidential and other
multistage election outcomes using state-level pre-election polls. The American
Statistician,62(1):1{10, 2008.
[9] Harry Enten. Congress's low approval rating is hurting republicans. Online,
-low-approvalrating-is-hurting-republicans/.
[10] The R Project for Statistical Computing. The r project for statistical computing. Online.URL
http://www [11] Andrew Gelman and Gary King. Why are american presidential election campaign polls so
variable when votes are so predictable? British Journal of Political Science, 23(04):409{451, 1993.
[12] Andrew Gelman, John B Carlin, Hal S Stern, and Donald B Rubin. Bayesian data analysis,
volume 2. Taylor & Francis, 2014.
[13] David Jarman. Daily kos election outlook: One of those times when we hate to be right. Online,
1342502/-Daily-Kos-Election-Outlook-One-of-those-times-when-we-hate-tobe-right.
[14] Drew A Linzer. Dynamic bayesian forecasting of presidential elections in the states. Journal of
the American Statistical Association, 108(501):124{134, 2013.19
[15] CN Morris. Comment. Journal of the American Statistical Association, 82(397):131{133, 1987.
[16] Manu Raju and John Bresnahan. How sen. hagans husband won stimulus cash. On-line,
-haganhusband-stimulus-cash-111339
[17] Douglas Rivers. Yougov poll performance in the 2014 senate elections. Online,
-pollperformance-2014-senate-elections/.
[18] Philip Rucker and Robert Costa. Battle for the senate: How the gop did it. Onfor-the-senate-how-the-gop-did-it/2014/11/04/a8df6f7a-62c7-11e4-bb14-
42
[19] Just Another Gibbs Sampler. Jags. Online. URL http://mcmc-
[20] Meredith Shiner. How the north carolina senate race became ground zero for
north-carolina-s-senate-race-became-everything-that-s-wrong-with-20119360
[21] Aaron Strauss. Florida or ohio? forecasting presidential state outcomes using reverse random
walks. In Princeton University Political Methodology Seminar, 2007.
[22] Karen Tumulty and Reid Wilson. Meet sean haugh, the libertarian pizza
guy who may deliver a senate seat in n.c. Online, July 2014. URL
-sean-haugh-the-libertarian-pizza-guy who-maydeliver-a-senate-seat-in-nc/2014/07/06/b321c03a-022f-11e4-8572[23] ESPN Internet
[24] Sam Wang. Exceptionally low turnout can account for polling errors. Online, Novem ber 2014.
-low- turnout-can-account-for-pollingerror/.
[25] Christopher Wlezien and Robert S Erikson. The horse race: What polls reveal as the election
campaign unfolds. International Journal of Public Opinion Research, 19(1)
43
Session 2: Exploring Economic Policy
45
Arming Women With Credit: Why Lending to Women Makes
Economic Sense by Victoria Monro
Final year, UCL Philosophy & Economics
Overall winner - Best Paper
and Presentation
Evidence suggests there is a structural issue, on a global scale, that affects women's ability to access
credit to the same degree as men and that this imbalance in credit availability has costs for welfare,
national and global economic growth. Much research comments on the effect on women from
financial empowerment (of which access to credit is one facet) - this paper seeks to consider the issue
from an economic perspective, rather than a social one, predominantly by focusing on the impact of
less credit availability on women's entrepreneurial abilities. In Section I, I indicate some of the ways
that women, face structural barriers to attaining credit. In Section II, I consider the impact of this lack
of access on economic growth, giving indication of the channels by which we miss out on the growth
premium of women's lack of access. In Section III, I offer a variety of motivations for policymakers to
make this a priority - ranging from the impact on national health and living standards to the impact on
productivity in agriculture and alleviating hunger and famine. Section IV focuses on the preconditions
of resolving the problem of gender discrimination in credit provision, be they legal or cultural
requirements to enable women to access credit. Section V concludes and summarises the main
findings.
Section I: Understanding the Difficulties Women Face Attaining Credit
Women struggle to access credit to the same degree as men, the world over. The concept of women's
financial liberation is often considered a developing-nation problem, but research indicates that
developed countries, including the United Kingdom, continue to face gender gaps in access to credit.
When it comes to attaining formal supplies of credit, women in developed nations face the bigger
credit gap (approximately four per cent). In developing nations, much credit is provided on an
informal basis, and the gender gap for such
types of credit is also in the region of four per
cent, indicating a global problem that manifests
in whichever type of loan is predominant in a
given country (Demirguc-Kunt, Klapper,
Singer; 2013). World Bank research notes that
businesses run by women tend to be less well
capitalised as compared with those run by men
in developing nations (Mason, King; 2001).
Further analysis by the International Financial
Corporation estimates the credit gap for women
in formal microenterprises in developing nations globally to be worth between $154-188 billion, for
small and medium-sized formal enterprises to be worth between $260-320 billion and for small and
medium-sized enterprises, both formal and informal, it is expected to lie between $750-920 billion
(2013). This implies credit issues grow worse as the businesses become more established.
A further study indicated that, in the USA technology industry, although there was no gender gap per
se in how much capital a start-up could attract, there were gender distinctions in how this capital was
attained, with women less likely to be able to access external credit (Robb, Coleman; 2009). The lack
of access to external sources of credit can restrict potential business development. The same study
found that women tended to start home-based businesses: to what extent this can be attributed to the
46
need to conserve funds raised privately because women recognise they are widely unable to access
credit, can only be hypothesised (Robb and Coleman note that home-based businesses in turn receive
less external funding - to what extent this is thus an endogenous factor to understanding the gender
gap has not been properly considered). The problem of credit availability is particularly acute in
agricultural farming in developing countries - for example, in Africa, where women have access to
just 1 per cent of the credit attributed to agriculture (The Montpellier Panel, 2012). The evidence, of
which this is just a selection, points towards a structural problem with women accessing loans: that it
is prevalent in all countries indicates that improving the relative wealth of a country is not sufficient
to enable women's financial liberation - and the subsequent economic gains - alone.
Section II: How More Credit For Women Can Improve Growth Prospects
Claims of economic gains from increased credit availability to women can be motivated via several
arguments, well grounded in empirical observations and studies. The first is a historic one: the impact
of wider employment opportunities for women has been huge economic gains - estimates of the effect
of narrowing the women-men employment gap suggest it is responsible for a quarter of European
gross domestic product growth since 1995 (OECD; 2008). Since progressing towards the elimination
of the gender gap in employment enabled these substantial gains in growth, it is plausible to suggest
growth can be attained from doing likewise when it comes to allocating loans.
If better use were made of women's human capital, global growth would increase at the same time as
global poverty would fall in every country (ibid.). If women have improved access to credit, they can
avail themselves of opportunities to exploit and improve their human capital, even when governments
do/change nothing. Women could use credit to improve education/technology/employment prospects
for them and their family, even in the absence of other types of government policies and programmes
to support them - thus, credit availability should be seen as at least one tool to make better use of the
skills of the world's women, even if it is wanted in an arsenal of many. This is reinforced by data that
demonstrates the clear gains that can be achieved if women are better educated: the more education a
woman gains, the lower is under-five child mortality, even after controlling for factors such as income
and socioeconomic status; children are more likely to be vaccinated if their mother is more educated
(Mason, King; 2001). A survey in Brazil showed a higher marginal return for spending on children if
additional income was given to the mother rather than the father (ibid.). Increased credit for women
can therefore be seen as a mechanism to increase human capital in the form of health and education,
via the increased spending on children it enables, and with this superior quality of human capital in
the future, the average economy will be able to do more with its resources - i.e. to achieve higher
levels of productivity, wealth. We can provide the theoretical framework that explains why we should
expect an expansion of an economy's production frontier due to better and more efficient labour, if
women can gain better access to external sources of funding - conversely, we see evidence that
permitting gender-based discrimination serves to reduce the human capital stock, and to
dampen/eradicate potential living standard improvements (i.e. World Bank, 2011). This should be
seen as a key example of how promoting women's entrepreneurship would have positive effects on
societal welfare.
The aforementioned gains are most commonly associated with developing nations. For example, we
would not expect significantly increased vaccination of children in developed nations such as the
United Kingdom, where vaccinations are the norm irrespective of women's financial empowerment,
thanks to a national health care system that ensures no costs are passed on to parents for the healthcare
requirements of under-18 year olds. An alternative account of the potential benefits to be accrued
needs to be provided; I turn to this now.
47
One of the main problems with a financial set-up that systematically disadvantages women is that it
creates an inefficient state of affairs, where resources are not being allocated to the most efficient
production method possible. Instead of women's great ideas being turned into viable, profit-making
businesses that create jobs, capital is allocated to worse ideas from men. Recent research based on
American technology entrepreneurs indicates that women are more capital-efficient than men,
creating businesses with less funding, and failing less often than average (Padnos, 2010). Given a
world of finite resources, and a post-crisis financial system that is still wary to lend, ensuring that
available credit is placed in the best possible hands is of importance to economic recovery, to
maximise returns from our limited resources and improve living standards.
In addition, research indicates that women with low-incomes are more likely to start businesses in
developed Europe than low-income men - which both reinforces the claim that increasing credit to
women could help reduce poverty, but also demonstrates that helping improve access to credit for
women is more likely to help the least-well off in developed society than corresponding loans to men.
Female entrepreneurs in developing European nations, despite being wealthier than average, perceive
fewer entrepreneurial opportunities than almost anywhere else in the world. Further, American
women that choose to exit their entrepreneurial endeavours cite financial issues twice as often as men
do (Kelley et. al; 2013). This indicates a large structural problem in developed nations which
reinforces the concern that the current credit distribution is too unequal, and inequitable, to constitute
the appropriate allocation of resources. Enabling women to access the credit lines they need, bearing
in mind the previously discussed evidence that indicates women can achieve higher returns with less
capital (see Padnos, 2013), would enable developed nations to become more productive, and to ensure
they were making the most of their talent pools. Not only would this result in higher output levels, but
this productivity gain, particularly when used to improve technology which drives growth could result
in further extension of the production frontier as innovations contribute to growth. Given the research
in technology shows that women are underfunded yet overachieving (ibid.), there are significant
potential economic gains from addressing the shortfall in credit attributed to women.
Section III: Motivating This Issue For Decision Makers
Previously cited evidence indicates the USA faces structural challenges in ensuring fair access to
women. However, if the UK were to merely match the USA's standards for credit with regards to
women, the UK economy would benefit from another 900,000 businesses and an extra £23 billion in
gross value added each year (Greater Return on Women's Enterprise; 2009). Furthermore, the UK
Government's Department for Business, Innovation and Skills indicates that the UK stands to create
as much as 150,000 extra businesses each year, if British women were creating businesses at the same
rate as British men. This corresponds to a loss in tax receipts and fewer jobs than would otherwise be
available (as women work for themselves, for example, they leave voids in companies that others
must fill, whilst potentially creating jobs within their new enterprise), as well as lost output. For
developed nations trying to promote growth in the wake of a severe financial crisis, this would make
some difference in aiding the economic recovery.
For developing nations, motivating for policymakers comes from research that demonstrates the
positive effects assisting women financially has on the stock of human capital in the economy,
ranging from improved health (via vaccinations, for example) to improved educational opportunities
for their children (Mason, King; 2001).
In addition, in developing nations, the economic gains estimated from creating a more equal
distribution of financial resource in households can be expected to be significant. The World Bank
48
estimates that more equitable control over resources and farming income within a household by
gender could result in farm yields being up to 20 per cent higher (2011). This result could have
significant positive consequences for nations that suffer from food instability.
For women who are able to receive their loans, studies in Bangladesh indicate that when women have
greater access to capital, their status and negotiating power in the household increases (World Bank;
2011) - in other words, the financial empowerment of women promotes the liberation of women more
generally. In some countries, this coupling of financial empowerment and liberation would be
undesirable; in the 11% of nations where women are legally obliged to obey their husband,
strengthening their bargaining position may not be a desired outcome (Demirguc-Kunt, A., Klapper,
L., Singer, D., 2013). For non-government organisations, for human rights groups and for charities,
the knowledge that programmes to improve women's access to loans have the ability to go beyond the
purely economic into the field of improving social justice, should be a powerful motive for action.
Importantly for fiscally-constrained governments, improving access to credit does not require that the
government stumps up the cash either - microcredit institutions are able to pick up some of the slack
required, continuing their successful trend. Data from 2006 indicates that microcredit institutions
made loans to more than 133 million people - 85% of the poorest of these people were women, with
hopes to reach 175 million in 2015 (Daley-Harris; 2006). Communications companies are becoming
innovative in how they ensure money gets delivered to rural areas - Safaricom's M-PESA in Kenya
enables people to credit their phones at an array of different locations with relative ease. 17 million
people use M-PESA for banking purposes - accounting for two thirds of the adult population (The
Economist; 2013). Creating the conditions for these programmes to flourish, legally and socially,
would enable other organisations to step in and fill the gap - in access to banking generally, as well as
the gender gap - to provide the loans, and not require costly government initiatives to provide credit
themselves. The OECD has previously said microfinance remains one of the most effective ways to
empower women (2008).
Importantly, women entrepreneurs are more likely to embark upon ventures that pursue social, as well
as economic, goals. Although at start-up the proportion of ventures that are committed to social goals
is equal across genders, ventures run by women are more likely to retain their focus on social
outcomes/goals by a further 9% (Meyskens, Allen, Brush; 2011). Research completed in the USA
indicates this result holds there too - women are more likely to start up social entrepreneurial ventures
(Van Ryzin et al; 2009). This implies, not just that women contribute to society via enhancing growth
and thus living standards, but that women are more likely to try to solve problems in their community
or society more generally. Enhancing women's ability to do this promotes social welfare beyond that
achieved by growth alone, and reduces the competition for government funds, allowing tax revenue to
be allocated elsewhere.
Section IV: The Conditions Required To Enhance Credit to Women
The conditions required to enhance credit to women depends on the starting position of the country in
question. It is necessary to distinguish between the conditions required of developing nations (which
are much more extensive and are more likely to require greater effort to achieve) and those of
developed nations (which are likely to require the government to confront biases and opaque decisionmaking processes on the part of creditors, particularly since gender-discrimination is already illegal in
many of these countries). I consider each in turn, beginning with developing nations.
For legal reasons, nations that do not allow women full property rights, or require women to defer to
male relatives, will struggle to advance women's access to credit. There is no incentive to lend to
someone who is legally obliged to do with the funds as another party sees fit, rather than to honour the
49
contractual agreements she herself has agreed to. A precondition for any discussion of women's
economic liberation - credit-based or otherwise - is their ability to control their economic status, as
workers, as earners, as borrowers and as savers. This is especially relevant in farming and agriculture.
Women produce 50% of all food, but unless they have strong access to, and control over, land, then
not only does productivity fall, but women struggle to access credit, as land is a vital source of
collateral (World Bank; 2001). Importantly, giving women the right to control their economic status is
insufficient if they cannot act upon its violation; in 11% of countries covered in the World Banks'
Women, Business and the Law report, women's testimonies in court are given lower weight (World
Bank; 2012a). In order for microfinancing companies to feel confident that the economic environment
supports their right to lend to women and will uphold their contracts with those women in the event of
violation (even against family members), such laws must be altered, or removed.
The second condition is that women must be seen as no less trustworthy for being women. In
Pakistan, for example, microfinancing requires two guarantors - neither can be female. Married
women must attain permission from her husband before she may take out a loan. Yet, estimates
indicate that when a woman does take out a loan, 50%-70% of the time the loan is utilised by male
relatives whilst the women continue to take responsibility for repayment (World Bank; 2012b).
Although this represents poor economic (and ethical) behaviour on the part of the micro-financiers,
the government has a role to play in ensuring that gender equality is taken seriously - particularly
since, as outlined, the economic gains can be so significant. Legislating against such practices, or
perhaps more simply, publicly encouraging microfinancing firms to recognise the value of women as
creditors and guarantors could be effective. Cultural and legal change may be necessary to promote
women's access to credit.
Another condition is that women must be able to accept control for assets and responsibility in the
household. When women are able to control their household's assets the likelihood of their using
formal financial products increases - based on results from 90 developing countries, 80% satisfy this
criterion, but another 20% stand to benefit from amending their laws (Demirguc-Kunt, Klapper,
Singer; 2013). This requires government action - either to remove or mitigate previous action they
have taken to undermine these requirements (i.e. having passed laws that obstruct women's ability to
own household assets), or to help promote the cultural change that would enable these to manifest.
Rich, OECD countries do not face these constraints in the way that many developing countries do or
might, but OECD countries have other challenges to face. For example, in addressing both the
perception that women-run businesses are less successful (when by many accounts, they're more
successful), and the perception that women are more likely to be unsuccessful in attaining formal
credit. We need to address whether or not women make decisions that result in receiving low levels of
credit as a consequence of anticipating low levels of credit, in order to ensure that women-run
businesses are not undercapitalised - for example, do women self-select into certain industries because
they perceive their chances of receiving the required level of credit is higher? If so, this indicates the
estimates of the gender-based credit gap could be grossly understimated. It is important that all
countries work to address this gap however it manifests - the gains may be more modest in developed
nations, but they are by no means inconsequential.
Section V: Concluding Remarks
This paper has demonstrated that there is a clear gender-based credit gap, and that this could be
underestimated if the existence of the credit gap is affecting women's choices of business. The
consequences of this gap include lower growth, lower human capital (education and health, primarily)
and lower living standards. Most importantly, the paper displays that whilst the problem is more acute
50
in developing nations, developed nations are not immune - relative wealth is no cure-all solution. This
should act to motivate NGOs, charities and policymakers to consider how they can best use their
resources to advance the economic fortunes of those most in need, and the economy as a whole. The
gains stand to be strong, if certain conditions are met, particularly those enshrining legal protections
for women akin to those enjoyed by men, and addressing perceptions that harm women's access to
credit in nations where such laws are already in place.
Bibliography
Daley-Harris, S. (2006). State of the Microcredit Summit Campaign Report 2004. Microcredit
Summit Campaign, Washington D.C.
Demirguc-Kunt, A., Klapper, L., Singer, D., 2013. Financial Inclusion and Legal Discrimination
Against
Women.
[pdf]
The
World
Bank.
Available
at:
<http://elibrary.worldbank.org/doi/pdf/10.1596/1813-9450-6416> [Accessed: 12 February 2015]
Department of Business, Innovation and Skills, 2011. Helping small firms start, grow and prosper.
[pdf] Department of Business, Innovation and Skills, UK Government. Available at:
<https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/32225/11-515-biggerbetter-business-helping-small-firms.pdf> [Accessed: 15 February 2015]
Economist Intelligence Unit, 2012. Women's Economic Opportunity 2012. [pdf] Economist
Intelligence
Unit.
Available
at:
<http://www.eiu.com/Handlers/WhitepaperHandler.ashx?fi=WEO_full_report_final.pdf&mode=wp&
campaignid=weoindex2012> [Accessed: 11 February 2015]
Hertz, N., 2011, Women and Banks: Are Female Customers Facing Discrimination? [pdf] Institute of
Public Policy Research. Available at: <http://www.ippr.org/publications/women-and-banks-arefemale-customers-facing-discrimination> [Accessed 23 January 2015]
International Finance Corporation, 2013. Access To Credit Among Micro, Small, And Medium
Enterprises.
[pdf]
International
Finance
Corporation.
Available
at:
<http://www.ifc.org/wps/wcm/connect/1f2c968041689903950bb79e78015671/AccessCreditMSMEBrochure-Final.pdf?MOD=AJPERES> [Accessed: 15 February 2015]
Kelley, D.J., Brush, C.G., Greene, P.G., Litovsky, Y., (2013). Global Entrepreneurship Monitor
2012 Women’s Report. Available at: <http://www.gemconsortium.org/docs/2825/gem-2012-womensreport> [Accessed: February 15 2015]
Klaser, S., 2003. In Search of the Holy Grail:How to Achieve Pro-Poor Growth? [pdf] Available at:
<http://www2.vwl.wiso.uni-goettingen.de/ibero/working_paper_neu/DB96.pdf>
[Accessed:
12
February 2015]
The Montpellier Panel, 2012. Women in Agriculture: farmers, mothers, innovators and educators.
London: Agriculture for Impact.
Mason, A., King, E., 2001. Engendering development through gender equality in rights, resources,
and voice. A World Bank policy research report. Washington DC; World Bank
Meyskens, M., Allen, E.I., Brush, C.G., 2011. Human Capital and Hybrid Ventures. Social and
Sustainable Entrepreneurship [published online]. Available at <http://dx.doi.org/10.1108/S10747540(2011)0000013007> [Accessed: 3 March 2015]
Moriah Meyskens, I. Elaine Allen, Candida G. Brush. "Human Capital and Hybrid Ventures" In
Social and Sustainable Entrepreneurship. Published online: 2011; 51-72.
51
OECD, 2008. Gender and Sustainable Development: Maximising the Economic, Social and
Environmental
Role
of
Women.
[pdf]
OECD.
Available
at:
<http://www.oecd.org/social/40881538.pdf> [Accessed: 11 February 2015]
Robb, A., Coleman, S., 2009. Sources of Financing for New Technology Firms: A Comparison by
Gender
[pdf]
Available
at:
<http://www.kauffman.org/~/media/kauffman_org/research%20reports%20and%20covers/2009/08/so
urces_of_financing_for_new_technology_firms.pdf> [Accessed: 15 February 2015]
Van Ryzin, G.G., Grossman, S., DiPadova-Stocks,L., Bergrud, E., 2009. Portrait of a Social
Entrepreneur: Statistical Evidence from a US Panel. Voluntas, 20(2), p129-140. Available at:
<http://link.springer.com/article/10.1007%2Fs11266-009-9081-4> [Accessed: 3 March 2015]
Women's Economic Task Force [WET], 2009. Greater Return on Women's Enterprise [pdf]
Available at: <http://www.womensenterprisetaskforce.co.uk/growe_report.html> [Accessed: 13
February 2015]
World Bank, 2012a. Women, Business and the Law 2012. Washington DC; World Bank [pdf].
Available at <http://wbl.worldbank.org/~/media/FPDKM/WBL/Documents/Reports/2012/WomenBusiness-and-the-Law-2012.pdf> [Accessed: 19 February 2015]
World Bank, 2012b. Are Pakistan’s Women Entrepreneurs Being Served by the Microfinance
Sector? [online] Available at <http://www.worldbank.org/en/news/feature/2012/10/17/are-pakistanswomen-entrepreneurs-being-served-by-the-microfinance-sector> [Accessed: 23 February 2015]
World Commission on the Social Dimension of Globalisation (WCSDG), 2014. A Fair
Globalisation: Creating Opportunities for All. [pdf] International Labour Organisation. Available at:
<http://www.ilo.org/public/english/wcsdg/docs/report.pdf> [Accessed: 12 February 2015]
52
Effect of the appreciation of the Swiss franc on the Ticinian Job
Market by Guido Tubaldi
Winner - Second Best Paper
and Presentation
2nd year, UCL Economics
On the 15th of January, the Swiss National Bank, decided to remove the cap on the Swiss Franc-Euro
exchange rate that was fixed at 1.2CHF/€ since 2011. I’m not going to look at the financial reasons
that have led the SNB to take this decision, but I’m going to investigate how the new exchange rate
will impact the labour market in Canton Ticino. I focus on this Swiss Canton, at the border with Italy,
comparing the results with other cantons that have similar economic features.
I’ve decided to look at this area because it allows me to carry out a natural experiment. I will study
the effect of the sudden exchange rate differential in a labour market characterized by three identical
groups of workers: Italian workers living in Italy, commuting on a daily basis in Switzerland; Italian
workers living and working in Lombardy; Swiss workers living and working in Ticino. The main
difference between these groups lies in their purchasing power.
I’m going to analyze the effect that this sudden exchange rate shock had on output and employment
levels in Ticino using the “3-equation model IS-PC-MR (Carlin et al. ,20143),” (without considering
the Central bank reaction to the shock). Ticino will actually react better to the loss in competition than
other Swiss cantons because of the more flexible nominal wages; Italian “Daily Cross Workers
Commuters” (DCWCS) will keep high their real wages even after having their nominal wages cut by
firms. Because there will be workers willing to work for lower nominal wages, the Italians, Ticinian
firms will have lower costs, and in this way they will not shut down. It’s clear that Italian workers will
be preferred to Swiss ones, at least in the manufacturing sector, which will be the most harmed by the
reduced competitiveness. The reduction in exports won’t be as dramatic and as long as in the other
regions where workers can’t cut their nominal wages without cutting their real wages. Even if there
will be a reduction in the numbers of Swiss people employed, because of their reduced employability,
many Ticinian producers will be able to survive thanks to cutting nominal wages of the Italian
DCWCSs, and in this way they will be able to save many other Ticinian jobs.
Integration of the Swiss Labour Market
Through the agreement on the free movement of people between Switzerland and the EU, signed in
1999 and effective since 2002, citizens coming from both areas, have obtained the right to choose
freely their place of work within the territories of the contracting parties. The free mobility treaty has
removed quotas in the labour markets and has removed the priority that native workers had in their
own nation. It meant that the labour market of the boarding regions, Lombardy in Italy and Canton
Ticino in Switzerland, became a unique and integrated labour market, as suggested by Maggi and
Gonzalez ( 2002).4
3
Carlin, Wendy, and David W. Soskice. Macroeconomics: Institutions, Instability, and the Financial System. Oxford University Press,
2014.
4
Gonzalez, Oscar, and Rico Maggi. "Segmentation by skills and wage discrimination in a trans-border labor market." ERSA
conference papers. No. ersa02p445. European Regional Science Association, 2002.
54
Figure 1: Ticino and Italian provinces of Como and Varese
The Italian provinces of Como and Varese, together with Canton Ticino can be though as “twins
separated at birth”. The latter, is now under the political and economic control of Switzerland, while
the are the twins” brought up by, Italy. Later in the paper, I will look at economic trend and it will
allow us to judge which one between the Italian twin and the Swiss one is the lucky one.
Across the border, people speak the same language, share the same traditions and have a similar
culture. The only true difference between the “twins” lays in the legislation. Therefore, workers
commuting and working on the other side of the border, don’t have to face any kind of cultural and
linguistic barrier that may negatively effect their employability and wage (According to Alberton and
Gonzalez, 2004 5 wages should converge to a common level in both side of the border). Any
difference we find in economic outcomes is not due to any personal disadvantage the worker faces in
the foreign labour market, rather it lays in the wider economic trends of Switzerland and Italy and,
consequently, of Canton Ticino and Lombardy. This is, in all effects, a natural experiment.
Economic trends
As the following data will clearly show, Lombardy and Ticino are characterized by similar Economy.
At the same time, even if we have similar economic features, we see that the Bilateral Agreement
results in migration singularly from South to North, i.e. from Lombardy to Ticino. For example,
looking at the growth trends of both areas since the signing of the Bilateral Agreement on the free
movement of workers in 2002, we can see a common path. Positive growth has been a constant fact
for both areas; this positive rate of growth for Canton Ticino is consistent with the Swiss
Confederation’s rate of growth, while Lombardy seems to be an exception in the Italian economic
situation.
Alberton, Siegfied, and Oscar Gonzalez. Monitoring a trans-border labour market in view of liberalization. Università della Svizzera
italiana, 2004.
5
55
Figure 26: Rate of growth for the 2 areas
Another relevant issue that needs to be analyzed is the level of unemployment. We can see how
Lombardy before 2009 (light green line), year in which the Italian economy as a whole suffered the
most from the global crisis, experienced considerably low levels of unemployment, which were in line
with the Swiss ones (blue line) and were almost 50% below the level we can see in Ticino (yellow
line). From 2009 there has been a rise in the level of unemployment for Lombardy and in 2011 it
overtook the level of Ticino. It’s important to see how the level of unemployment in Ticino and in
Lombardy, sharply rise in 2011, differently from what happens in the rest of the Swiss Confederation,
where the level of unemployment is constant on the pre-crisis level.
Figure 37: Unemployment levels in Lombardy (green) , Switzerland (blue) and Ticino(yellow)
Another important fact that we need to stress is the almost equal division of the labour force by
sectors of employment. In the provinces of Como and Varese, there are almost 250,000 workers in the
secondary sector and more than 380,000 in the tertiary sector. Quite similar percentages, with clearly
different absolute number due to different size of population, are applied to Canton Ticino, as the
following graphs show.
6
7
Swiss Data taken from USTAT, Lombardy Data taken from “Annuario statistico regionale”
Data on employment for Ticino and Lombardy come from USTAT
56
Figure 48: Occupation by sector in Como&Varese and Ticino
These common economic trends in the two areas, and the complete openness of the two markets, may
lead us to think that the market clears; people would move across the two sides of the border till they
are indifferent on where to work. In reality, as previously stated, the migration is just from one side;
from Lombardy to Ticino. It is the phenomenon of the DCWCSs, also called the “frontalieri”.
“Frontalieri” - DCWCS, for salary reasons
Their status as daily DCWCSs is recognized and regulated by special laws between Italy and
Switzerland. In order to be subjected to this particular legislation, that regulate aspects like taxations,
pension schemes and unemployment benefits, you need to live no farther than 20 kms from the Swiss
border. In fact, as I’ve previously written, there are only Italian “frontalieri”. Since the Bilateral
agreement became effective in 2002, the number of the “frontalieri” has doubled, from 30000 units to
more than 60000 at the end of 2014 and they now account for more than ¼ of the whole Ticinian
Labour force, as shown by the following graphs.
Figure 59 : Number of “Frontalieri” and Ticinian Labour Market
The macroeconomic data I’ve previously reported are not enough to explain this phenomenon. There
must be something else that encourages every day more than 60,000 people to drive for so many
kilometres and make them the most envied workers in the area. It is the Swiss Franc and the different
8
9
Data for Varese comes from “Annuario statistico regionale”, Data for Ticino come from USTAT
Data on the number of “frontalieri” and on the division of labour by nationality come from USTAT 2013
57
purchasing power you have by spending a Swiss income in the relatively cheaper side of the border,
the Italian side.
Envied by Italian workers working in Italy
By looking at nominal income, we see that there’s an impressive difference between the average
nominal income earned by workers in Varese or Como and the average nominal income workers
would get in Ticino.
If we consider, as an average nominal monthly income for an Italian worker in the manufacturing
sector in Varese, the estimate of 2.150€, equivalent to 2800CHF (assuming an exchange rate
1.3CHF/€), the following graphs, helps us understanding how the “frontalieri” are advantaged
compared to Italians.
Figure 610: Wages by qualification levels and by working permit
Looking at the previous graph, we can see how the “frontalieri” have an immense advantage over
their colleagues working in Italy. Even the least qualified DCWCSs , manage to have a considerably
higher salary than Italian workers.
(Partially) Envied by Swiss workers too
If nominal wages for DCWCS are actually higher than the nominal wages they would get in Italy, we
need to look at real wages to compare Ticinian workers with the “frontalieri”. What really matters is
the purchasing power workers have.
Swiss
CHF
€11
100
76.92
Lombardy 75
57.69
The table shows that in order to buy a basket of goods in Ticino you need 100CHF.
The same basket of goods in Italy costs 75CHF. It means that life in Italy is ¼ cheaper than in
Switzerland.
10
11
Data on average nominal income by qualification and work permit come from USTAT 2011
According to an average exchange rate for 2011 equal to 1.3CHF/1€
58
At the same time, data (figure 5) show that the nominal income of Italian commuters is sensibly lower
than nominal income of natives, and this actually compensates for the purchasing power gap. This
higher nominal income is a double edged situation for Swiss workers, since they manage to keep good
standards of living in the more expensive Ticino, but at the same time it could make them less
employable, since their income results in higher costs for business. It contributes to create a situation
of wage dumping, where the cheaper labour supply from immigrants causes native’s wages to fall.
Figure 712: Ticinian labour market by sector and by nationality of workers
For these reasons, we have seen how the Italian DCWCS have been taking a bigger and bigger share
of the secondary sector, where their general lower level of education doesn’t harm their employability
chances.
It has also had political implications in Ticino, where nationalist parties, worried about the possibility
that DCWCSs’ pressure on both unemployment and wages, are getting more and more consensus. In
February 2014, the result of a consultative referendum in Switzerland, about the possibility of reimposing quotas on the number of foreign workers in the Swiss labour market, had the highest
percentage of “yes” vote (68.2%13 of the voting population) in Canton Ticino, although Ticino is not
the only Swiss Canton that is experiencing the phenomenon of cross border commuting.(In both
French and German speaking parts of the Swiss Confederation, there’s even a higher number of daily
DCWCS)
Exchange rate shock
Since 2011 the Swiss National Bank decided to peg the value of its currency to the Euro at the fixed
exchange rate of 1.2CHF/1€. It created stability and certainty both in financial and labour markets. On
the 15th January 2014 the SNB suddenly decided to remove the peg allowing the exchange rate to
fluctuate. Immediately, the Swiss Franc appreciated against the Euro, peaking at 0.85CHF/1€, before
stabilizing at 1 CHF/1€. People, living at the border between Italy and Ticino, rushed to change their
money, buying the relatively depreciated Euros and spend them in Italy. This exchange rate shock had
an immediate effect on the competitiveness of Switzerland:
Qο€½
12
13
p*e
p
Data on the division of labour by sector come by USTAT 2011
Swiss referendum on 9th February 2014: http://www.swissinfo.ch/eng/ticino-says--basta---to-cross-border-workers/37943558
59
where Q is the real exchange rate, reflecting the competitiveness of a country, e is the nominal
exchange rate (CHF/1€), p* is the Italian price level and p is the Swiss price level. The Swiss Franc
has appreciated against the Euro, therefore e ο‚― , reducing Q, and the Swiss competitiveness too. For
these reasons the Swiss economy is now out of the Trade Balance curve; it’s importing more than
exporting. This process can be represented as a shift along the Aggregate Demand (curve) on the 3
equation model, AD-BT-ERU;
AD represents aggregate demand for home’s good (in this case Swiss goods),
BT curve represent the balance of trade
ERU curves represent the equilibrium on
Figure 8: Shift along AD, from A to B, competitiveness goes down
This movement along the AD curve, due to e ο‚― , has also effects on the labour market; it causes an
upward shift of the PS curve, since real wages actually go up, due to the increased purchasing power
workers now have. At the same time, nominal wages are at B’, on the WS curve.
(N.B: the distance between BC is equal to the distance between B’C’)
Figure 9: PS curve shifts up
60
Both the difference between real wage at C’, on PS’, and the nominal wage at B’ and the lower level
of employment, N1, generate deflation. This can also be told by the fact that B is on the left of the
ERU curve that pins a condition in which inflation is constant. It creates a downward trend for Swiss
deflation, that is already low (-0.3%). Inflation moves from point A’’ to point B’’. Because output is
still out of the equilibrium point Ye and expectations are assumed to be adaptive, inflation will go
down till the economy is back to an equilibrium.
Figure 10: Effect on inflation due to the differential between WS and PS
Over multiple periods of time, deflation has an effect on both nominal wages and prices. As we have
previously stated, Q ο€½
p*e
. Then, since p ο‚― , Q starts rising, and the economy slowly starts moving
p
from B back to A (figure 11). It means that we slowly re-acquire competitiveness through lowering
prices and adjusting nominal wages. Moreover, the increase in Q pushes PS’ downwards, till it will
lay on PS.
As previously stated, inflation will go down further, since the economy is still below the equilibrium
output. The closer the economy gets to the equilibrium point, the smaller the decrease in inflation will
be.
This is the theoretical path the economy should follow, but we have to consider that the extremely
prolonged time of both high unemployment and low nominal wages (being on the left of the
equilibrium point), may create a Hysteresis effect in the market. Workers are not happy of seeing their
nominal wages cut and therefore they are less inclined to work. At the same time, firms close
permanently since their expectation for the future are extremely negative because of the appreciated
currency that makes them less competitive; the process of regaining competitiveness, through nominal
wage and price adjustment, takes too much time. It means that many firms close permanently, fire
employees and consequently reduce output.
61
Figure 11: Increase in Q due to adjustments in nominal prices
Both the cut in nominal wages and the loss of jobs due to many firms shutting down, have a negative
effect in the labour market; it creates a high number of long term unemployed individuals that are not
willing to supply their effort for lower nominal wages. Being unemployed for a long time leads to a
progressive loss of skills and to an erosion of psychological attachment to working life. It makes
unemployed people not to push enough pressure in the labour market. As a result, we have a higher
WS curve, because of the lower pressure of unemployed on employed people. A shift of the WS
curve, creates a new point of equilibrium in the supply side of the economy, making the ERU curve
shift leftwards too. At the same time, the fact that many firms will actually shut down, leaving people
unemployed has a negative impact on the AD. The reduced capacity of exporting of Swiss firms,
results in a negative impact on the BT curve too. Both of them, in the end, will shift leftwards.
62
Figure 12: Hysteresis in the Swiss market.
For Ticino, the situation is different. The Ticinian job market will react differently since it is much
more flexible because of the DCWCS. As data previously showed, they are much more willing to
work for lower nominal wages, because their real wages manage to be high thanks to the lower cost of
living. Differently from their Swiss colleagues, they care much more for their real wage than for the
nominal wage, and therefore are not discouraged by the cuts in nominal wages. Moreover an
appreciated Swiss Franc, allows the “frontalieri”, to have a higher purchasing power, even higher than
before.
Swiss
CHF
€14
€15
100
76.92
100
57.69
75
Lombardy 75
If before, with the exchange rate at 1.3CHF/1€ Italian commuting in Switzerland needed 57.69 € to
buy a basket of good in Italy, now, with the exchange rate 1CHF/1€, they have 75€, that are more
14
15
I consider an average Exchange rate for 2011 equal to 1.3CHF/1€
I consider an Exchange rate for 2015 equal to 1CHF/1€
63
than the 57.69€ they need for the same basket. Italians are even more willing to supply their work,
putting pressure on Swiss workers. As a result, in Ticino the WS curve won’t shift at all because the
DCWCs will keep a high pressure on local workers. It means that also the ERU curve will be were it
used to be.
Ticinian firms will substitute from Native Ticinian workers to DCWCs since they accept receiving a
lower nominal wage. It allows a rapid adjustment of prices and nominal wages and it results in a
faster return to the equilibrium point, avoiding the Hysteresis effect (Ticinian economy follows
exactly the theoretical pact of Figure 11). The flexibility given by the “frontalieri” has an effect also
on the demand side of the economy because it reduces labour costs for firms that will manage to
avoid the big loss in competitiveness that other Swiss firms may face. Companies, instead of firing
and closing, will assume mainly DCWCS, reducing their costs. Reduced costs mean more funds
available for investments, therefore the AD curve instead of shifting leftwards due the lower
competitiveness, will remain where it currently is16. (Figure 12; blue colour for other cantons where
there’s no possibility of hiring DCWCs, black colour for the Swiss situation of equilibrium before the
shock and for Ticino after and before the shock).
Figure 12: Ticino reaction compared to similar cantons’ reaction.
As we can see from figure 12, the Ticinian market ends up with the same level of employment and
output. It reacts better than the less flexible labour markets we find in other Cantons, where both
output and employment are at lower levels (Point A’). The flexibility given by the presence of the
Clearly we can’t generalize for the whole Swiss market, but we can compare Ticino to other cantons that have common economic
features.
16
64
“frontalieri” often seen as having a negative impact on the whole labour market, is actually the reason
why Ticino would react better. It’s true that DCWCSs will substitute natives or, as the nationalist
Ticinian propaganda prefers to say, “steal natives’ jobs”. However thanks to them the effect of the
shock on the whole economy is attenuated. It’s possible that avoiding closing other firms, have saved
more Swiss jobs than the ones that actually are “stolen” by Italian workers.
Conclusions
With this paper I’ve tried to stress how the presence of these Italian daily DCWCS is actually making
the Ticinian labour market more flexible and more able to react to this exchange rate shock. Their
presence actually allows to limit the loss in competition that Swiss companies would have from a
relatively stronger Swiss Franc against the Euro.
The fact that they are willing to work for lower nominal wages, since their real wages are unchanged
because of the higher purchasing power they have in Italy, allows firms to cut costs and survive,
avoiding the Hysteresis effect. Nationalist propaganda will probably look at the fact that the much
more employable Italians will steal Swiss jobs in this period of lower output due to lower exports; but
I think that through both my application of this model and the use of data, in the future we will be able
to see whether the presence of the “frontalieri” has saved more Swiss jobs than it has stolen.
This will actually allow for a more sane discussion on the effect of migration in Ticino based more on
actual economic data and less on slogans.
References:
Alberton, Siegfied, and Oscar Gonzalez. Monitoring a trans-border labour market in view of
liberalization. Università della Svizzera italiana, 2004.
Carlin, Wendy, and David W. Soskice. Macroeconomics: Institutions, Instability, and the Financial
System. Oxford University Press, 2014.
Gonzalez, Oscar. Wage differential of a trans–border labor market, a quantitative analysis. Diss.
University of Lugano, 2007.
Gonzalez, Oscar, and Rico Maggi. "Segmentation by skills and wage discrimination in a trans-border
labor market." ERSA conference papers. No. ersa02p445. European Regional Science Association,
2002.
Nerb, Gernot, et al. "Scientific report on the mobility of cross-border workers within the EU27/EEA/EFTA countries." Final report. European Commission DG Employment and Social Affairs
Brussels (2009).
65
Localised complementary currencies: the new tool for
policymakers? The Sardex exchange system
2nd year, UCL Economics and Philosophy
“Unconventional times call for unconventional remedies.
That is the lesson we have learnt since the crisis.” Martin Wolf
Abstract
Sardex was created in 2010 in Sardinia, Italy. Today, it has 2500 users and it is growing
exponentially. It is one of the three pilot projects of the European Union’s Digipay4Growth, which
tests three complementary currencies’ capacity to provide business and access to credit to SMEs and
foster growth: in Sardinia, Bristol, and Catalunya.
This paper analyses the microeconomic benefits for Sardex users, building the macroeconomic
framework necessary to model the Sardex network. It evaluates the advantages to consumers, firms,
and policymakers, adapting the models of investment (Tobin’s q) and consumption (permanent
income hypothesis) underlying the traditional IS curve, to the dual currency framework. It shows that
Sardex increases consumption and investment, enabling firms to obtain higher profits, and enabling
consumers to smoothen consumption by relaxing credit constraints, while supporting the local
economy. Hence, it shows that Sardex produces a permanent real output gain, a countercyclical,
stabilising effect on the economy, and that it provides additional sources of income for the
government - besides an extraordinary economic insight into the economy.
Having refuted the claims for which it harms the primary currency, or that it constitutes a financial
threat, the conclusion that this type of system should be implemented by policymakers is reached.
Introduction
This essay argues that governments should use a regional complementary exchange system with
interest free loans to aid achieving sustainable local growth.
Firstly, the paper introduces the complementary exchange system that will be examined throughout
the paper: the “Sardex” exchange network.
Secondly, it analyses the microeconomic benefits for Sardex users, hence modelling their
macroeconomic behaviour. Firms face lower marginal cost of capital and higher demand. Tobin’s q is
consequently adapted to the new framework. Consumers face reduced credit constraints, and support
local communities with regional-focused spending. The effect of relaxed credit constraints is shown
on the permanent income hypothesis.
Thirdly, the paper discusses Sardex’s benefits for policymakers, namely: a permanent real output
gain, a stabilising, countercyclical effect on real output, an additional source of income, and an
extraordinary insight into regional economic dynamics.
Ultimately, the paper discusses the claims for which Sardex may harm Euros’ circulation, or
constitute a financial threat. Having shown that risks are managed and contained, the paper concludes
that policymakers should implement complementary currencies to achieve their goals of sustainable
local growth.
66
The Sardex model
Agents pay a small fee to access the Sardex network; there, firms buy goods by creating interest-free
debits. Debits have to respect a debt-ceiling, proportional to each firm’s productive capacity, and have
to be cleared within a set time frame. Buyers’ debits correspond to sellers’ credits; the latter can be
spent immediately. Debits are cleared when debit-holders sell goods and services, hence gaining
credits. Therefore, goods are paid through past or future provision of goods to other agents (Sardex srl
2015).
Hence, the supply of money is self-adjusting, and is equal to the economy’s aggregate credits (or
debits). If n is the number of indebted users i, and N-n is the number of non-indebted users j,
𝑛
𝑠
𝑀𝑆𝑅𝐷
𝑁−𝑛
= ∑ 𝑑𝑒𝑏𝑖𝑑𝑖 = ∑ π‘π‘Ÿπ‘’π‘‘π‘–π‘‘π‘—
𝑖
(1)
𝑗
Given the Fisher relation and zero nominal interest, the real interest rate is:
π‘Ÿπ‘†π‘…π· = 0 − πœ‹π‘†π‘…π·
(2)
Where πœ‹π‘†π‘…π· is Sardex-inflation. Companies are obliged by the terms of the contract to sell at the same
price at which they would sell in Euros. (Sardex srl 2015) Since most companies in the network sell in
both currencies, enforcement of this rule is effective: SRD1 (the unit of account) buys what €1 would
buy:
πœ‹€ = πœ‹π‘†π‘…π·
(3)
and
π‘Ÿπ‘†π‘…π· = −πœ‹€
(2.1)
Because SRD credits/debits only arise from exchanging of goods within the network, one cannot buy
and sell SRD on Forex markets. The result is a separate economy where large volumes of liquidity
can circulate, despite stagnation and lack of “ordinary” money.
The next section analyses formally why optimising agents use Sardex, and what the macroeconomic
implications of their behaviour are.
Microeconomic benefits, and macroeconomic implications
We will consider two types of agents: firms, and consumers, maximising profits and
utility respectively. We assume that Sardex’s access fee is negligible, and that one can
readily enter the network if they want to.
Firms
Microeconomic behaviour
Profit maximising firms always choose the cost-minimising inputs vector. (Cowell, 26). We assume
no qualitative difference between capital purchased in Sardex, and in Euros. Assuming that marginal
cost of capital, π‘€πΆπ‘˜ , is given by r, the real interest rate, plus the depreciation rate 𝛿 and given
π‘Ÿ€ > π‘Ÿπ‘†π‘…π· , we will have that
π‘Ÿ€ + 𝛿 > π‘Ÿπ‘†π‘…π· + 𝛿
⇔ π‘€πΆπ‘˜ 𝑆𝑅𝐷 > π‘€πΆπ‘˜ €
By 4.1, capital will be cheaper in Sardex. We infer:
67
(4)
(4.1)
Lemma 1. Cost-minimising firms always obtain capital from SRD, rather than Euros, until the
borrowing limit (𝐷𝑆𝑅𝐷 ) is reached.
Hence, if dSRD is Sardex-debt,
π‘€πΆπ‘˜ 𝑆𝑅𝐷
π‘€πΆπ‘˜ = {
π‘€πΆπ‘˜ €
𝑖𝑓 𝑑𝑆𝑅𝐷 < 𝐷𝑆𝑅𝐷
(5)
𝑖𝑓 𝑑𝑆𝑅𝐷 = 𝐷𝑆𝑅𝐷
𝛿 − πœ‹€
𝑖𝑓 𝑑𝑆𝑅𝐷 < 𝐷𝑆𝑅𝐷
𝑖. 𝑒. π‘€πΆπ‘˜ = {
𝛿 + 𝑖 − πœ‹€
𝑖𝑓 𝑑𝑆𝑅𝐷 = 𝐷𝑆𝑅𝐷
(5.1)
Notice that here i is the interest rate that firms would obtain from commercial banks (Carlin, 160),
which is a mark-up on the Central Bank’s rate, so that π‘Ÿ€ > π‘Ÿπ‘†π‘…π· even in times of low Central Bank
rates.
Macroeconomic behaviour: investment and “Tobin’s q”.
Consider a representative firm. Define q as the ratio between the marginal benefit of capital, and the
marginal cost of capital. (Carlin, 29)
π‘ž=
π‘€π΅π‘˜
π‘€πΆπ‘˜
(6)
π‘€π΅π‘˜ is the effect of capital on total revenues.
π‘€π΅π‘˜ =
=(
πœ•π‘‡π‘… πœ•π‘‡π‘… πœ•π‘„ πœ•(𝑝 ∗ 𝑄)
=
=
π‘€π‘ƒπ‘˜
πœ•π‘˜
πœ•π‘„ πœ•π‘˜
πœ•π‘„
πœ•π‘
πœ•π‘ 𝑄
𝑄 + 𝑝) π‘€π‘ƒπ‘˜ = 𝑝 (
+ 1) π‘€π‘ƒπ‘˜
πœ•π‘„
πœ•π‘„ 𝑝
1
= 𝑝 (1 + ) π‘€π‘ƒπ‘˜
πœ‚
(7)
Where p is price, Q is output, πœ‚ is elasticity of demand, and π‘€π‘ƒπ‘˜ is marginal product of capital (we
assume
πœ•π‘€π‘ƒπ‘˜
πœ•π‘˜
< 0 to capture diminishing marginal returns). Plugging (5.1) and (7) into (6),
π‘€π΅π‘˜
π‘€πΆπ‘˜ 𝑆𝑅𝐷
𝑖𝑓 𝑑𝑆𝑅𝐷 < 𝐷𝑆𝑅𝐷
π‘ž=
(8)
π‘€π΅π‘˜
{π‘€πΆπ‘˜ €
𝑖𝑓 𝑑𝑆𝑅𝐷 = 𝐷𝑆𝑅𝐷
1
𝑝 (1 + πœ‚ ) π‘€π‘ƒπ‘˜
𝑖𝑓 𝑑𝑆𝑅𝐷 < 𝐷𝑆𝑅𝐷
𝛿 − πœ‹€
𝑖. 𝑒.
π‘ž=
(8.1)
1
𝑝 (1 + πœ‚ ) π‘€π‘ƒπ‘˜
{ 𝛿 + 𝑖 − πœ‹€
𝑖𝑓 𝑑𝑆𝑅𝐷 = 𝐷𝑆𝑅𝐷
Investment will increase if q>1, and will decrease if q<1. If q>1 implied a level of capital associated
with a Sardex debt greater than 𝐷𝑆𝑅𝐷 , companies would invest in Euros until q=1.
68
If we did not have the Sardex system, we would have π‘ž =
denominator, so q should increase. As we assumed
πœ•π‘€π‘ƒπ‘˜
πœ•π‘˜
1
πœ‚
𝑝(1+ )π‘€π‘ƒπ‘˜
𝛿+𝑖−πœ‹€
. Eliminating i lowers the
< 0, agents will invest, lowering π‘€π‘ƒπ‘˜ , until
q=1. Lemma 2 follows.
Lemma 2. Investment always increases if the Sardex exchange network is introduced.
We now follow a similar procedure for consumers.
Consumers
Microeconomic behaviour
Let us consider the utility-maximising quantity of a good x, supplied in both Euros (π‘₯€ ) and Sardex
(π‘₯𝑆𝑅𝐷 ), so that
π‘₯𝑆𝑅𝐷 + π‘₯€ = π‘₯
(9)
Agents choose π‘₯𝑆𝑅𝐷 and π‘₯€ maximising U(π‘₯𝑆𝑅𝐷 , π‘₯€ ), subject to π‘₯𝑆𝑅𝐷 + π‘₯€ = π‘₯. Because agents may
potentially purchase only from one system and prices are equal, π‘₯𝑆𝑅𝐷 and π‘₯€ will be perfect
substitutes:
U(π‘₯𝑆𝑅𝐷 , π‘₯€ ) = aπ‘₯€ + 𝑏π‘₯𝑆𝑅𝐷
(10)
If Sardex products were more time-consuming to source than euro-products, it may be that b<a. Then,
individuals would only consume π‘₯€ . They would only consume π‘₯𝑆𝑅𝐷 if they wanted to sustain the
local economy: b>a. Consumers would be indifferent if a=b. This is shown in the diagram below.
Solutions are at P, Q, and on the constraint.
Studies have shown that households tend to prefer consuming in local complementary currencies,
because money remains within the local network to support the economy. (Hoffman, 4-5) This means
that for a considerable amount of Sardex users, we should expect b>a. This is consistent with Littera,
Panadyotis, Dini, and Sartori’s analysis of Sardex. (Littera, 12)
Lemma 3: consumers can sustain the local economy using Sardex.
69
To understand consumer-benefits further, we use Friedman’s Permanent Income Hypothesis
(henceforth: PIH). To simplify the algebra, let us assume that individuals live over an infinite lifespan.
Supposing individuals are not credit constrained, and that they maximise the present value of the
utilities U(𝑐𝑑 )=log(𝑐𝑑 ) (which display diminishing marginal returns), PIH holds that
𝑐𝑑 =
π‘Ÿ
Ψ𝑒
1+π‘Ÿ 𝑑
(11)
where
∞
Ψ𝑑𝑒
= (1 + π‘Ÿ)𝐴𝑑−1 + ∑
𝑖=0
1
𝑦𝑒
(1 + π‘Ÿ)𝑖 𝑑+𝑖
(12)
𝑒
which is the present value of expected lifetime wealth at time t: 𝐴𝑑 is assets at time t, and 𝑦𝑑+𝑖
is
expected after-tax income in period t+i (Carlin, 37). Optimal consumption is then smooth across the
lifecycle, and predictable swings of the business cycle. When individuals earn less than they want to
consume, they borrow. When they earn more, they repay the debts, and/or save money. (Ibid)
Thus, access to credit is essential for smooth consumption. The next section analyses the effect of
introducing Sardex in a scenario where individuals are credit constrained, i.e. they cannot borrow as
much as wish to.
Macroeconomic effect of relaxing credit constraints
Let us suppose, to simplify the algebra, that individuals’ income was made up of an infinite stream of
equal payments. Then,
∞
π‘Ÿ
1
𝑐𝑑 =
𝑦̅] = 𝑦̅
[(1 + π‘Ÿ) ∗ 0 + ∑
(1 + π‘Ÿ)𝑖
1+π‘Ÿ
(11.1)
𝑖=0
Suppose that at t=2, consumers were announced an increase in income at t=3. Agents would borrow at
t=2, so that they could partially anticipate future earnings and consume more at t=2, where marginal
utility is higher because of lower consumption (blue dotted line below). However, if agents were
completely credit constrained, consumption would keep tracking income, jumping at t=3 (red line).
70
We could think of agents who are partially credit constrained as lying between the blue and the red
line. By relaxing credit constraints, Sardex enables individuals to get closer to the blue optimal line,
anticipating some of the increase in income (green dotted line), in periods with higher marginal utility
because of lower consumption. The more constraints are relaxed, the more consumers redistribute
future changes, the smoother consumption will be.
Lemmas 4 and 5 follow.
Lemma 4: Sardex enables credit constrained agents to get closer to their optimal consumption.
Lemma 5: Sardex stabilises consumption: the increase is smaller, and more gradual.
The next section analyses the policymaking implications of Lemmas 1-5.
Advantages for the policymaker
Advantage 1. Permanent output gain
An IS curve represents the demand-side of the economy. It displays the inverse relationship between
interest rates and real output. Underlying the IS curve, there are microeconomic models of optimal
71
consumption (PIH) and investment (Tobin’s q) (Carlin, 16). Let us consider the IS curve for the total
economy. Let us temporarily assume that agents in Sardex operate in the same framework as agents in
Euros. The total IS curve will be the horizontal addition of ISSRD and IS€. Suppose rmin is the lowest
rate the Euro-economy can get to, given banking markups on central bank rate.
Once we introduce the zero nominal SRD rate, we move along the ISSRD curve, leading to higher
output 𝑦2 𝑆𝑅𝐷 . As agents can now borrow in Sardex, consumption and investment will increase for any
given r (relaxation of quantity credit-constrain). This leads to 𝑦3 𝑆𝑅𝐷 .
Proposition 1. The Sardex system produces a permanent real output gain.
72
Advantage 2. Stabilising effects
The quantity theory of money asserts that in a given time period, if M is the amount of money in the
economy, V is the velocity of circulation (how many times M is used), P is the price level, and T is
the volume of transactions, MV=PT must hold. (Naqvi, 3) Substituting T with real output as is
common in economic literature,
𝑀𝑆𝑅𝐷 𝑉𝑆𝑅𝐷 = π‘ƒπ‘Œπ‘†π‘…π·
(12)
Where π‘ƒπ‘Œπ‘†π‘…π· is nominal Sardex-GDP. Because the complementary currency is small relatively to the
Euro-region and prices are equal, we assume P to be exogenously determined by the Euro-Economy.
In a recession, we expect credit constraints to rise: individuals have lower income, and risk premia
rise. (Iacoviello, 764) By lemma 5, PIH agents who were not credit constrained before, will move to
Sardex to smoothen consumption. Then, as demand for goods rises, sellers should join the Sardex
network to seize additional profit opportunities.
Two different effects should then take place. As more agents engage in the economy taking more
debt, 𝑀𝑆𝑅𝐷 rises, hence increasing output, given constant 𝑉𝑆𝑅𝐷 . As more consumers crowd the
marketplace without actively creating debt themselves, 𝑉𝑆𝑅𝐷 will increase, because output circulation
will increase. Given 𝑀𝑆𝑅𝐷 𝑉𝑆𝑅𝐷 = π‘ƒπ‘Œπ‘†π‘…π· , the latter enhances the effect of the former. The final effect
can be represented as a rightward shift of the 𝐼𝑆𝑆𝑅𝐷 curve.
Lemma 6: the Sardex economy mitigates the fall in total income in a recession.
On the other hand, when the Euro-economy grows, credit constraints should fall: individuals have
higher income, and risk premia decrease. Agents who are still credit constrained would keep using
73
Sardex debits. But agents that now attain their smooth consumption level, would stop borrowing, and
would pay back debts. Hence, 𝑀𝑆𝑅𝐷 falls, and the Sardex economy shrinks.
The overall effect can be represented as a leftward shift of the 𝐼𝑆𝑆𝑅𝐷 curve.
Lemma 7: the Sardex economy mitigates the fall in total income in a boom.
Authors have pointed out that complementary currencies’ countercyclical benefits are do not
significantly impact the bigger nation. (Krohn, 2008). Yet, this does not imply that they may not make
a significant difference locally, in the weaker parts of the economy. Stodder has shown that Swiss Wir
turnover (a bigger equivalente of Sardex, established in 1934) is directly correlated with
unemployment.
(Stodder, 15). His empirical conclusions may be criticised because his paper does not account for
endogeneity issues: turnover may be determined by interest rate policies, rather than unemployment.
74
However, it could be argued that his findings seem to fit our theoretical framework and to corroborate
Sardex’s counter-cyclicality.
From lemmas 6 and 7, we infer:
Proposition 2: Sardex has a countercyclical, stabilising effect on the local economy
Advantage 3. Additional sources of income
We analyse two ways to finance government spending through Sardex: cheaper debt, and
monetisation of productive capacity.
Cheaper debt
Let us suppose that the Government could borrow money from the Sardex system, paying it back
through goods and services. This is currently not the case in Sardinia, where the government does not
have access to Sardex borrowing. We could model debt dynamics with
βˆ†π‘ = 𝑑 + (π‘Ÿ − 𝛾)𝑏
where βˆ†b is the change in debt-to-GDP ratio, b is debt-to-GDP ratio, d is existing debt, r is the interest
rate, and 𝛾 is the growth rate of the economy (all in real terms). If π‘Ÿ − 𝛾 ≥ 0, βˆ†π‘ increases, and future
debt rises. If π‘Ÿ − 𝛾 < 0, future debt-to-GDP ratios will fall (Carlin, 520).
In a recession, 𝛾€ < 0, while we expect the Sardex economy to grow: 𝛾𝑆𝑅𝐷 > 0. Given π‘Ÿπ‘†π‘…π· = −πœ‹,
we would have that
−πœ‹ − 𝛾𝑆𝑅𝐷 < (𝑖 − πœ‹) − 𝛾€
i.e.
π‘Ÿπ‘†π‘…π· − 𝛾𝑆𝑅𝐷 < π‘Ÿ€ − 𝛾€
From which we infer:
Proposition 3. Sardex-debt is cheaper than Euro-debt in recessions
Proposition 3 arises from lower real Sardex interest rates, and higher Sardex growth. Notice that if the
economy were not deflating significantly, Sardex debt-to-GDP should also fall:
−πœ‹ − 𝛾𝑆𝑅𝐷 < 0
Monetisation of unused productive capacity
Lower interest rates, higher Tobin’s q, and lower credit constraints enable the Government to benefit
from additional aggregate demand in Sardex, just like any other seller. If demand for goods sold by
the government is πœ’(π‘ž, πœ‰, π‘Ÿ), where q is tobin’s q, πœ‰ is the level of credit constraints, and 𝛼 is some
positive constant, we have that
πœ’(π‘ž, πœ‰, 𝑖 − πœ‹) − πœ’(π‘ž, πœ‰ − 𝛼, −πœ‹) = βˆ†πœ’
Where βˆ†πœ’ is the amount of goods that the Government can sell in Sardex, but not in Euros. Selling βˆ†πœ’
in Sardex provides the government with extra revenue, without creating additional government debt.
Policymakers could use these Sardex credits to pay public employees, or for Helicopter money stimuli
– which Sardinian institutions already considered (Regione Sardegna, 2015).
When governments sell βˆ†πœ’ and receive credits, they receive an amount of credits corresponding to an
increase in 𝑀𝑆𝑅𝐷 . These can be reassigned to agents for purposes of output targeting. Given MV=PT,
75
if Sardex velocity is higher than Euro velocity, then it will be more efficient to use helicopter stimuli
in Sardex, rather than Euros. Current measurements by the Sardex company give 𝑉𝑆𝑅𝐷 = 12.28 and
𝑉€ = 1.5 (Sardex srl, 2015). Thus, we infer:
Proposition 4: If 𝑉𝑆𝑅𝐷 > 𝑉€ , and βˆ†πœ’ > 0, the government can stimulate the Sardex economy
without expanding its debt.
The following section outlines the “informational” benefits provided by the Sardex system, explaining
how they can be used to optimise government stimuli’s efficiency.
Advantage 4. Knowledge of economic structure
The economy can be seen as a “directed network” of nodes, in which “directed links” between nodes
(non-symmentric connections) represent economic transactions. The money stock would follow the
directed links, and GDP would then be the sum of the “walks” of the money stock in the network
(Jackson, 511-520).
This implies that given different structures, the money stock will flow differently, and GDP dynamics
will be different. Below, the liquidity in subnetwork N keeps circulating from A to E in a cycle, while
in N’ it leaves the subnetwork. If velocities were different, given the same money stock, GDP would
be different.
Thus, a policymaker who planned to inject liquidity in the system should account for the structure of
the economic network to maximise the stimulus’ efficiency. The same reasoning applies to
government spending.
In the Euro Economy, it is extremely complex for policymakers to account for these dynamics. How
to best spread the effect of an helicopter money stimulus? Which taxes have the strongest impact on
money circulation? The Sardex company, has perfect knowledge of the entire spending network.
Hence, it is in a better position to to answer these questions. Governments could share data with the
company and use it to monitor the economy, tailor their policies to the network structure, and readily
measure their efficacy.
Proposition 5: Governments policies can be optimised and tailored to the network structure, in the
Sardex system.
Given propositions 1-5, this paper concludes that there are compelling arguments for policymakers to
use Sardex to aid achieving goals of local sustainable growth.
76
The next section answers two claims against this conclusion.
Assessing concerns
Someone could argue that Sardex may harm Euros’ local circulation, and that 𝑉𝑆𝑅𝐷 > 𝑉€ is a result of
“Gresham’s law”, for which “bad money” (less useful Sardex-credits) drives out of circulation “good
money” (more useful Euros) (Mundell, 1998). Hence, it may be argued that policymakers should limit
Sardex’s expansion.
However, Sardex requires users to pay sums above a certain threshold (which depends on the user’s
characteristics) in both Euros, and Sardex (Sardex, 2015). That already ensures that Sardex is a
complement, and not a substitute, of the Euro. Moreover, even if such requirement were not in place,
the efficiency gains from trade would provide incentives to Sardex-users to engage in transactions
with non-Sardex users – locally, Nationally, and internationally. And because Sardex cannot be
exchanged on foreign markets, agents would have to hold euros.
Someone may reply that Sardex exposes the economy to financial instability threats: it enables credit
constrained agents with high default risks to take on new loans. Such view, however, would disregard
the fact that Sardex assesses default risk (lower because of the lower rate), by evaluating the
productive capacity of new users and the demand for their products. The company assigns the
maximum debit level so that the default risk is contained. Therefore, risk is managed by Sardex, and
the argument is rejected.
It has to be emphasised, however, that if policymakers thought that Sardex were a threat, the
network’s size could readily be manipulated. A cap on its number of members would limit growth.
The average individual debit ceiling could be legally reduced, limiting risk and turnover. Taxes could
be enforced, so that the marginal cost of capital in Sardex rises, and cost-minimising firms don’t find
it optimal anymore to invest through SRD. Thus, risks could be managed by the Government, too.
Conclusions
This paper has introduced the dynamics of the Sardex system. It has shown that firms maximise
profits through the lower marginal cost of capital (Lemma 1), and that investment will be increased if
Sardex is introduced (Lemma 2). It has also shown that credit constrained consumers can support the
local economy (Lemma 3) while smoothening their consumption and getting closer to their optimum
(Lemmas 4 and 5). Hence, it has shown that Sardex produces a permanent output gain, and a
countercyclical, stabilising effect on real output (propositions 1 and 2). It has also shown that Sardex
provides cheaper ways to finance government spending, and a vehicle of more effective demand-side
stimuli, as we all as an extraordinary economic insight into the economy (propositions 3-5). It has
concluded that these are compelling arguments for the system’s implementation, and it has rejected
the arguments for which Sardex harms the primary currency, or constitutes a financial threat.
“Unconventional times call for unconventional remedies.” This essay has argued that the Sardex
model may be one of them.
77
Bibliography
Cowell, Frank. Microeconomics - principles and analysis. Oxford: Oxford University Press, 2005.
European Commission. Digipay4Growth. 2014. http://www.digipay4growth.eu (retrieved on February
25, 2015).
Gregory A. Krohn, Alan M. Snyder. «An economic analyseis of contemporary local currencies in the
United States.» International Journal of Community Currency Research 12 (2008): 53-68.
Hoffman, Torsten. «How can Local Currencies Enhance Economic Activity in Communities? .»
Oxford Business school, 2010.
Iacoviello, Matteo. «House Prices, Borrowing Constraints, and Monetary Policy in the Business Cycle
.» The American Economic Review 95 (2005): 739-764.
Jackson, Matthew O. «An Overview of Social Networks and Economic Applications .» Handbook of
Social Economics 1 (2011): 511-585.
Mona Naqvi, James Southgate. «Banknotes, local currencies and central bank objectives .» Quarterly
Bulletin 2013 Q4, Bank of England, 2013.
Mundell, Robert. «Uses and Abuses of Gresham's Law in the History of Money.» Columbia
university . 1998. http://www.columbia.edu/~ram15/grash.html (consultato il giorno February 26,
2015).
Regione Autonoma della Sardegna. «Moneta complementare: Cappellacci, reddito comunità con
circuito Sardex .» Regione Autonoma della Sardegna. 2015.
http://www.regione.sardegna.it/j/v/25?s=225092&v=2&c=241&t=1 (consultato il giorno Februarly
14, 2015).
Sardex srl. «Sardex Press Kit.» 2015.
Stodder, James. «Complementary Credit networks and Macro-Economic Stability: Switzerland’s
Wirtschaftsring .» Journal of Economic Behavior & Organization 72 (October 2009): 79–95.
Wendy Carlin, David Soskice. Macroeconomics - Institutions, intstability, and the Financial System.
Oxford: Oxford University Press, 2015.
Wir Bank. Wir Bank. 2015. http://www.wir.ch/it/la-banca-wir/ (consultato il giorno March 2, 2015).
78
Savings-Linked, Saving Lives: The Role of Conditional Cash Transfers
in Financial Inclusion by Isaac Lim
Final year, UCL Economics
Introduction
Academics, governments and concerned parents alike have long espoused the importance of saving.
Yet low-income individuals and families (especially in developing countries) face barriers to saving
that have resulted in saving rates which are far below optimal. In this regard, this paper aims to
reinforce the case for linking savings to conditional cash transfers (CCTs) as a means of improving
saving rates. This paper also highlights 4 key design features that are necessary for the pivot towards
increasing saving rates or financial inclusion through CCTs.
Savings among the poor has taken on many forms, both formal and informal, as the poor resort to
second-best mechanisms in the absence of affordable and viable alternatives. Recent studies have
shown that even in urban cities like Bogota, Columbia and Mexico City, Mexico, between 65-85% of
households do not hold formal savings accounts.17 This percentage is likely to be even higher in rural
areas where accessibility to such banking facilities pales in comparison.
On the other hand, CCT programmes worldwide provide monetary assistance to eligible low-income
families. Transfer payments typically increase with the number of children in a family. Redemptions
of transfers are conditioned on the completion of certain actions, often associated with education and
health-related goals. There is however, still scope for expanding CCTs to include increased saving
rates as an additional objective.
In this paper, section 1 explains why savings are important for the poor, and weighs the pros and cons
of formal and informal saving mechanisms. Section 2 highlights the barriers to formal saving for lowincome people that desperately need to be addressed. Lastly, Section 3 states the case for linking
savings to CCTs and proposes 4 key design features that should be included in the marriage of the two
concepts, namely: 1) technology; 2) choice architecture; 3) financial education; and 4) government.
Section 1: Why Are Savings Important For The Poor?
Empirical studies have long since dispelled the myth that the poor are simply too poor to save and
instead spend all their income on consumption. In fact, Rutherford (1999) insists that “the poor want
to save, and do save”.18 In this section, we explore different types of saving and the savings options
available to low-income individuals.
A. Types of savings
Saving and its benefits can be dichotomised into two main types. First, saving can refer to storing
small sums of money to smooth day-to-day consumption. The poor are often faced with unstable
incomes and may regularly encounter periods of low or zero cash inflows and need to store money for
daily use.
17
18
Solo (2008)
Rutherford (1999)
80
Financial Diaries true case study:
Hamid is a reserve driver of a motorised rickshaw in Dhaka. His unpredictable income is
determined by how many hours he can work (if at all, since he is a reserve), how many
times the rickshaw breaks down and how many customers he ferries on a daily basis.
Collins, Morduch, Rutherford and Ruthven (2009)
Second, savings may be the accumulation of larger lump sums of money for investments 19 ,
spending on festivals/“life events”20, and self-insurance against emergencies21. For many individuals
and families, this aspect of saving is the first step towards breaking free from the poverty cycle.
B. Formal and Informal Savings Mechanisms
Low-income savers also face the choice between engaging in informal or formal savings mechanisms.
Informal mechanisms include rotating savings and credit associations (ROSCAs), credit
cooperatives22, pay-to-save programmes, etc., where money is usually stored in cash at home, with
informal deposit collectors or with other individuals in the community. Conversely, formal savings
mechanisms involve storing money in formal accounts with established institutions like banks and
microfinance institutions. In the following table, we see a breakdown of the main pros and cons for
both informal and formal savings mechanisms.
Formal Savings Mechanisms
Informal Savings Mechanisms
Low accessibility
Easy access
Individuals who store money under their mattresses or flowerpots can reach for it as and when they
like, whereas those who store money in formal savings accounts must withdraw money from the bank
or ATM, which can be extremely inconvenient.
Greater deposit security
Vulnerable to theft and embezzlement.
Aside from the exposure to theft, individuals also face possible situations where their deposit
collectors and/or fellow members of savings clubs, who are likely to be susceptible to the same the
negative aggregate shocks, may choose to abscond with the money when savers need it the most. 23
Reduced peer-pressured lending
Susceptible to peer-pressured lending
With the privacy accorded through formal savings mechanisms, needy friends and family members
are less likely to have knowledge of the total amount saved, thus individuals may be less pressured
into lending otherwise conspicuous and readily available sums of money.24
Lower cost of saving
High cost of saving
Despite their minimum deposit/balance requirements and other non-price barriers, formal savings
mechanisms may still cost less for the saver, in comparison to the high fees charged by informal
deposit collectors25 or the opportunity costs of attending regular 1-2 hour ROSCA meetings.26
Table 1. Pros and Cons of Formal and Informal Savings Mechanisms
19
Banerjee and Duflo (2007)
Ibid
21
Collins et al. (2009), Krishna (2003)
22
Armendaríz and Murdoch (2011)
23
Townsend (1994)
24
Kendall (2010)
25
Rutherford (1999)
26
Aryeetey and Gockel (1991)
20
81
Clearly the advantages of engaging in formal saving outweigh its disadvantages. Indeed, when
presented with formal savings opportunities, the poor have time and again demonstrated their
eagerness to save through these channels and have benefitted as a result.
There are a number of experiments that corroborate these claims. In a randomised control trial in
western Kenya, Dupas and Robinson (2013a) witnessed an 87% uptake among those who were
offered a chance to set up a savings account without the imposition of a US$10 minimum balance.
About 4-6 months after opening savings accounts, women in the treatment group were reportedly
investing 38-56% more in their businesses daily, and enjoying and 37% higher personal expenditures
than women in the control group.
Section 2: Barriers to Microsaving
When we refer to formal savings mechanisms for low-income families, we inevitably refer to
microsaving, which is a component of microfinance that involves incentivising poor individuals to
save in small deposit accounts. 27 Recognising the importance and benefits of engaging in formal
savings mechanisms as stated in the previous section, we shall explore the main barriers that forestall
the proliferation of microsaving from both consumer and provider perspectives.
Consumer
Provider
Lack of physical access leading to high costs:
High transaction costs:
ο‚· Consumers must not only pay for
ο‚· Microsavings are typified by
transport to the deposit/withdrawal point,
small,
infrequent
but must also bear the opportunity cost of
transactions.
work hours sacrificed.
ο‚· In
a
study
of
61
ο‚· Beneficiaries of the Bono de Desarrollo
microfinance institutions in
Humano (BDH) CCT programme in
Latin America, operating
Ecuador, are estimated to face opportunity
costs for balances lower
costs of about US$0.40 per hour, or almost
than US$100 were found to
half of the US$1.00 minimum wage in
be between 200-300% of
Ecuador in 2004.28
the amount of savings
High direct costs of transactions:
mobilised
from
these
accounts.29
ο‚· Direct costs including minimum balance
ο‚· These
operating
costs
requirements, and maintenance and
consist of costs “related to
withdrawal fees.
personnel (teller services),
ο‚· E.g. Solo (2008) cited that 70% of the
communications, security,
Mexicans and 65% of the Colombians
use of equipment and
interviewed in his study “stated that
electricity,
and
other
banking fees, required minimum balances
30
expenses”.
and initial deposits were simply too high
ο‚· Microsavings however, may
for them to pay.”
grow over time; low-income
Bad treatment by and mistrust of financial institutions:
savers may not remain small
27
http://www.investopedia.com/terms/m/microsavings.asp
Carrillo and Jarrin (2007)
Portocarrero, Tarazona and Westley (2006)
30
Portocarero, Tarazona and Westley (2006)
28
29
82
ο‚·
ο‚·
Fear of rejection by banks and general
distrust of private financial institutions is
an emotional variable that has been
addressed through various methods.31
For example to offset consumers’
reluctance to store deposits with formal
financial institutions, some governments
have engaged in funding deposit insurance
programmes.
savers forever, and thus
could
potentially
be
profitable in the long run.
Table 2. Barriers to Microsaving
Section 3: Why and how should we use Conditional Cash Transfers (CCTs)
to encourage saving?
We have explored how savings are an essential element of poverty alleviation and identified the main
barriers to microsaving. We now seek to propose a solution that comprehensively addresses these
obstacles. This section attempts to redefine the rationale behind integrating savings with CCTs and
proposes four essential components that should be included in any savings-linked CCT programme
design framework.
Linking savings to CCT programmes has already been explored in a number of CCT programmes
worldwide. By taking advantage of the nature of and infrastructure provided by existing CCT
programmes, governments and other stakeholders may circumvent the aforementioned barriers,
thereby arriving at an effective and cost-efficient solution to the current lack of financial inclusion
among the poor.
Ideally, achieving altruistic health and education objectives should remain the primary focus of CCTs,
but we must also acknowledge that societies stand to reap much larger social dividends through the
inclusion of a savings component within these programmes. Rather than detracting from the original
CCT objectives, linking savings to CCTs may well provide an avenue for beneficiaries to use their
transfers more prudently and thus escape poverty more effectively.
CCTs present extremely attractive savings opportunities for a number of reasons:
1) Self-selection: CCTs are targeted at the poorest of the poor. The eligibility requirements for
CCT payments ensure that the impact of linking savings to CCTs will be concentrated on
CCT beneficiaries – those who need it the most and are least likely to engage in formal
saving.
2) Low marginal cost: For each transfer payment, a transaction between the CCT providers and
recipients is already expected regardless of the amount that changes hands. Even if CCT
beneficiaries choose to withdraw a smaller amount of the transfer and save the rest, the CCT
provider will incur minimal or no additional cost for that particular transaction.
3) Information: Information on individuals and families can be used to generate customised
expense allocation recommendations for the beneficiaries. For the most basic allocation of
resources (to categories such as education, food and gas), only simple observations such as
the number of school-going children and the number of people in the household are needed.
31
Solo (2008)
83
These data points are likely to be readily available through the targeting mechanisms used by
the CCTs.
4) Existing conditions: As CCTs are conditioned on various behaviours, encouraging financial
education through these programmes will be easier. Foundational financial literacy-related
topics (such as how to use savings accounts, Automated Teller Machines (ATMs), etc.) could
be included in classes that beneficiaries are already required to attend as part of the CCT.
Further to these points, this paper proposes 4 key components to be included when designing savingslinked CCTs, specifically technology, choice architecture, financial education, and government.
A. Technology
Graph 1: % Reduction in Cost through Electronic Payments
100%
83%
80%
62%
60%
40%
20%
0%
Brazil
South Africa
The rapid advancements of new technologies in the area of branchless and mobile banking have
already greatly reduced transaction costs and operating overheads, relative to the costs associated with
traditional bricks-and-mortar bank branches. For example, through the use of electronic payments, the
administration costs of the Bolsa Familia CCT programme in Brazil fell to 2.5% of the payment
value, down from 14.7% originally.32 Other CCT programmes worldwide have experienced similar
cost reductions along with other benefits such as lower corruption, leading to higher amounts of
grants reaching their intended recipients.33
While the ATM is certainly a viable option for streamlining payments delivery and may be the most
conventional substitute for a traditional bank branch, there are still fairly high sunk costs associated
with installing physical ATMs. Under the JUNTOS CCT programme in Peru, 67% of beneficiaries
already receive payments through savings accounts with the state bank, Banco de la Nación, but there
are still neither banks nor ATMs in 80% of districts where beneficiaries reside as of Dec 2014. 34 In
fact, JUNTOS beneficiaries spend on average an estimated 5 hours to reach the nearest ATM or bank
branch, incurring around 10% of the payment value in transportation costs. 35 Facing similar
constraints, governments have turned to alternative methods of payment delivery including mobile
phones and point-of-sale (POS) devices.
32
(Savings-Linked Conditional Cash Transfers: Lessons, Challenges & Directions, 2011)
Ibid
Ayling (2014)
35
Ibid
33
34
84
Although banking through mobile phones and POS devices require trained agents to carry out
transactions, this should by no means be perceived as a limitation. Instead, the versatility afforded in
the choice of agents (typically grocery stores, lottery booths, pharmacies, etc.) allows for maximum
reach and scalability of this venture. In rural areas, where constructing and maintaining an ATM may
not be as cost effective, agents who are equipped with the right skillset can disburse cash and accept
deposits with minimal interference with their daily operations and at similarly low costs.
Mobile phone banking in particular, has gained popularity in some developing countries and taps on
existing infrastructure; most residents (even in developing countries) have access to a mobile phone,
with mobile penetration rates reaching 70% in India, 69% in China and 55% in Africa.36 Through the
M-PESA programme in Kenya, telecommunications provider Safaricom built up a network of MPESA agents who were trained to accept and dispense cash in exchange for the transfer of an equal
value of mobile credits (an easily accessible store of value for users). 37 Cash collection and restocking
are carried out by “master agents” daily.38
Case Study:
In November 2008, the Oportunidades programme in Mexico entered into a partnership
with Diconsa, the largest retail store chain in Mexico, where stores were provided with a
POS device and employees were trained to disburse CCT payments through these devices.
Transaction costs fell from 30.1 pesos to 0.49 pesos and opportunity costs for beneficiaries
shrank from 16.9 pesos to 2.2 pesos per transaction.1 Diconsa stores that provided this
service also experienced a 20-30% increase in sales compared to control stores.
Savings-Linked Conditional Cash Transfers: Lesson, Challenges and Directions (2011)
With minimal disruption, governments can work with agents and private institutions to disburse CCT
payments through these means. These platforms offer increased accessibility for beneficiaries and
reduced transaction costs for providers that could engender an increase in formal savings rates.
Platforms alone however, do not offer much added incentive to save. This brings us to the next
component which is choice architecture.
B. Choice Architecture
In The New Microfinance Handbook for the World Bank, Ignacio Mas finds that poor people often
earmark their assets for different expenses. For example, “the goats are to pay for school fees and
uniforms, [and] the rotating savings and credit association account is to buy a sewing machine”. If
formal savings could be fragmented, individuals can be encouraged to include savings as part of their
income allocation. Each payment withdrawal is a window of opportunity to influence beneficiaries’
savings decisions.
The term “choice architecture”39 describes the nuanced influencing of the choices of a target audience.
In this situation, choice architecture may be used to subtly encourage savings through the use of the
calculated positioning of options on ATM, POS devices and mobile phone screens. Currently, many
ATMs are programmed to dispense cash as simply as possible, with generic default amount options
presented to users. Similarly, on mobile phones installed with mobile applications such as M-PESA,
users can go through a number of steps using the application to craft a message that will be sent to the
service provider, which will then execute the transfer of mobile credits.
36 http://blogs.worldbank.org/category/tags/mobile-phone-penetration
37 Mas and Morawczynski (2009)
38 Ibid
39 Nudge: Improving Decisions About Health, Wealth and Happiness. Thaler and Sunstein (2008)
85
Instead of keeping the cash withdrawal as stylised as possible, this paper recommends adding in an
additional step before individuals select an amount to withdraw, which will present beneficiaries with
a recommended breakdown of how their income should be allocated. On mobile phone platforms for
example, individuals may choose from one or more of the pre-assigned values or choose to withdraw
a custom amount through an additional step. In his paper on mobile money management, Ignacio Mas
suggests an alternative system of deferred payments, whereby consumers can set dates on which they
transfer money to themselves. These delayed transfers correspond to specific purposes, like
purchasing a bicycle, in essence “saving up” a certain amount for future expenditure.40 This would
correspond to the “Set Target” option on the Services menu shown in Figure 1.
While these nudges do not necessarily prevent the individual from withdrawing the total sum of their
transfer, there are two channels through which this additional step could possibly have a positive
impact on savings among users. Firstly, the additional step makes withdrawing the total amount just
slightly more troublesome, creating a weak psychological barrier associated with withdrawing the
entire transfer sum. Secondly, by using information taken from the CCT household surveys and
devising a portfolio allocation of sorts, individuals no longer have to informally earmark their assets
according to expected expenditures but can rely on electronically generated allocation. Individuals
should also be allowed to set financial goals or customise options for themselves by updating their
profiles at the bank or through these platforms.
C. Financial Education
Aside from raising awareness and reducing distrust41, financial education should aim empower CCT
beneficiaries to take full advantage of the range of services provided. While most may have mobile
phones or access to ATMs/POS equipped stores for example, they may not know how to use these
platforms to access their savings accounts, reducing the effectiveness of mobile banking platforms. In
a survey conducted on the Bono de Desarrollo Humano program in Ecuador, it was found that only
39.4% of beneficiaries rated their ability to use their newly issued ATM cards as well or very well.42
40
Mas (2007); Mas and Mayer (2011)
41 Maldonado, Moreno-Sanchez, Perez and Orjuela (2011)
42 Samaniego and Tejerina (2010)
86
Individuals may find the influx of information overwhelming, and therefore it will be important to
teach financial literacy in easily digestible stages. Financial literacy can be introduced initially at the
most fundamental levels, which would include teaching people how to bank using ATMs and their
mobile phones applications. This could be accomplished through short videos or individual training,
as has been done by Familias en Accion in Colombia as part of their Plan de Bancarizacion. Once
people are familiar with these basic functions, they could then learn about the importance and
incentives of formal saving. When savings-related knowledge has been taught, individuals could then
continue to learn about the use of other financial products (e.g. loans) that are offered by banks and
microfinance institutions (MFIs). Over time, as beneficiaries become more comfortable with formal
financial products, financial inclusion is likely to rise in tandem.
Financial education can be disseminated through channels such as advertising or even included as a
condition for receiving the CCT payment itself. Past experience has shown the effectiveness of raising
schooling rates among children from poor families; there is no reason why this same result could not
apply to financial literacy among adults as well.
D. Government
The government will play a broad role in ensuring the success of such a monumental increase in
financial access for the poor. Banks and MFIs are unlikely to be profitable from the start due to the
sudden increase in micro-savings accounts with low initial deposit values. Although these accounts
may eventually grow large enough to be self-sustaining, the cost of setting up such an account may be
too large a burden for private enterprises to bear.43 Intuitively, the presence of positive externalities
from introducing technology and access to the low-income CCT beneficiaries indicates that such
facilities will not be provided for under free market conditions by private enterprises alone. Thus the
cost of opening and maintaining these accounts must be at least subsidised by the government. If
possible, to increase trust among beneficiaries of private financial institutions, governments could also
provide a measure of “deposit insurance”.
The government can also help to build infrastructure, develop technology and foster conditions
favourable to banks and MFIs dealing with micro-savings. Yet it must also remain cautious that
actions taken to ease the burden of MFIs do not erode competitiveness and efficiency. Taking
experience from past successes and through government-to-government sharing, governments may
also be able to lend technical assistance to banks and MFIs providing micro-savings facilities. Lastly,
financial education for the poor, who may be unable to afford professional courses, may also require
government funding.
Conclusion
The design features presented above are aimed to address the barriers to microsaving directly.
Implementation of technology and strong government support should reduce transaction costs for the
microsavings accounts, which will hopefully translate into lower costs of opening and maintaining
formal savings accounts as well as increase the accessibility of such services for the poor. Choice
architecture and financial education concurrently tackle issues including mistrust of the private
financial sector and the lack of financial literary among the poor.
Savings-linked CCTs however, are a fairly new concept and there is still a dearth of empirical
evidence. Much ground has to be covered if we wish to ascertain a positive causal relation between
linking savings to CCTs and an increase in savings. Correspondingly, there may remain other
43 Seibel (2005)
87
obstacles that have yet to surface. As such, flexibility will be essential in implementing a dynamic
solution to address low saving rates. Nevertheless, I hope that the core ideas expressed in this paper
serve as a useful tool when exploring the possibility of savings-linked CCTs in the future.
References
Ananth, B., Chen, G., & Rasmussen, S. (2012). The Pursuit of Complete Financial Inclusion: The
KFGS Model in India. Consultative Group to Assist the Poor/The World Bank.
Aportela, F. (1999). Effects of Financial Access on Savings by Low-Income People. Banco de
Mexico, Research Department.
Armendaríz, B., & Morduch, J. (2010). The Economics of Microfinance Second Edition. Cambridge,
Massachusetts: The MIT Press.
Aryeetey, E., & Gockel, F. (1991). Mobilizing Domestic Resources for Capital Formation in Ghana:
The Role of Informal Financial Sectors. Nairobi: African Economic Research Consortium, Initiatives
Publishers.
Ashraf, N., Karlan, D., & Yin, W. (2006). Tying Odysseus to the mast: Evidence from a commitment
savings product in the Philippines. Quarterly Journal of Economics , 673-697.
Ayling, S. (2014, December 16). CGAP. Retrieved March 3, 2015 from Can Agents Improve
Conditional Cash Transfers in Peru?: http://www.cgap.org/blog/can-agents-improve-conditional-cashtransfers-peru
Banerjee, A., & Duflo, E. (2007). The Economics Lives of the Poor. Journal of Economic
Perspectives , 141-167.
Boulding, K. E. (1970). Economics as a Science.
Bouman, F. (1977). Indigenous savings and credit societies in the developing world. Savings and
Development , 181-214.
Carrillo, P., & Jarrin, J. (2007). Efficient Delivery of Cash Transfers to the Poor: Improving the
Design of a Conditional Cash Transfer Program in Ecuador. Institute for International Economic
Policy.
Collins, D., Morduch, J., Rutherford, S., & Ruthven, O. (2009). Portfolios of the Poor: How the
World's Poor Live on $2 a Day. Princeton: Princeton University Press.
Collins, D., Morduch, J., Rutherford, S., & Ruthven, O. (2009). Portfolios of the Poor: How the
World's Poor Live on $2 a Day. Princeton: Princeton University Press.
De Los Rio, J., & Trivelli, C. (2011). Savings Mobilization in Conditional Cash Transfer Programs:
Seeking Mid-term impacts. Social Protection for Social Justice. Institute of Development Studies,
UK.
Deaton, A. (1989). Saving in Developing Countries: Theory and Review. The World Bank Annual
Conference on Development Economics (pp. 61-96). The International Bank for Reconstruction and
Development/The World Bank.
Dupas, P., & Robinson, J. (2013a). Savings Constraints and Microenterprise Development: Evidence
from a Field Experiment in Kenya. American Economic Journal: Applied Economics , 163-192.
Edgeworth, F. (1925). The pure theory of monopoly. Collected Papers relating to Political Economy ,
1.
88
Friedman, D. (n.d.). The World According to Coase. University of Chicago Law School Record .
Hindricks, J., & Myles, G. D. (2013). Intermediate Public Economics (2nd Edition ed.). Cambridge,
Massachusetts: The MIT Press.
Jenkins, C. (2014, October 10). Lecture 2 Competition Analysis. ECON3012 - Industrial Economics I
. UCL, London.
Kendall, J. (2010). A Penny Saved: How Do Savings Accounts Help the Poor? Seattle: Bill and
Melinda Gates Foundation.
Krishna, A. (2003). Falling in Poverty: Other Side of Poverty Reduction. Economic and the Polical
Weekly , 533-542.
Maldonado, J. H., & Terjerina, L. (2010). Investing in large scale financial inclusion: The Case of
Colombia. Inter-American Development Bank.
Maldonado, J. H., Moreno-Sanchez, R., Perez, I. G., & Orjuela, C. A. (2011). Conditional Transfer
and Financial Inclusion Programs: Opportunities and Challenges in Latin America. Proyecto Capital.
Mas, I. (2013). Beyond Products: Building Intergrated Customer Experiences on Mobile Phones. In J.
Ledgerwood, The New Microfinance Handbook: A Financial Market System Perspective (pp. 299318). The World Bank.
Mas, I., & Mayer, C. (2011). Savings as Forward Payments: Innovations on Mobile Money Platforms.
Mas, I., & Morawcyznski, O. (2009). Designing Mobile Money Services: Lessons from M-PESA.
Innovations , Vol. 4, No. 2.
Nash, J. (1950). Equilibrium points in n-person games. Retrieved 2014 from Proceedings of the
National
Academy
of
Sciences
of
the
United
States
of
America:
http://www.pnas.org/content/36/1/48.full
Pickens, M., Porteous, D., & Rotman, S. (2009). Banking the Poor via G2P Payments. Washington
D.C.: Consutative Group to Assist the Poor.
Portocarrero, F. M., Tarazona, A. S., & Westley, G. D. (2006). How Should Microfinance Institutions
Best Fund Themselves? Washington, D.C.: Inter-American Development Bank.
Rutherford, S. (1999). The Poor and Their Money: An essay about financial services for poor people.
Department for International Development .
Samaniego, P., & Tejerina, L. (2010). Financial Inclusion Through the Bono de Desarrollo Humano
in Ecuador. Inter-American Development Bank.
(2011). Savings-Linked Conditional Cash Transfers: Lessons, Challenges & Directions. New York,
NY: Report from a Global Colloquium.
Seibel, H. (2005). The Microbanking Division of Bank Rakyat Indonesia: A Flagship of Rural
Microfinance in Asia. Working Paper . University of Cologne, Development Research Center.
Smith, S. (2015, February 3). Lecture 4 - Externalities. ECON7005 - Economics of the Public Sector .
UCL London.
Solo, T. M. (2008). Financial exclusion in Latin America - or the social costs of not banking the urban
poor. Environment and Urbanization , 47-67.
Steel, W. F., & Aryeetey, E. (1994). Informal savings collectors in Ghana: Can they intermediate?
Finance and Development , 36-37.
89
Townsend, R. M. (1994). Risk and Insurance in Village India. Econometrica , 539-591.
Varian, H. (2006). Intermediate microeconomics: a modern approach (7 ed.). W. W. Norton &
Company.
Yunus, M. (2002). Grameen Bank II: Designed to open new possibilities. Dhaka: Grameen Bank.
90
Session 3: Deeper exploration through an
undergrad dissertation
91
Breaking up Germany: A review of the drivers of social mobility by
Dennis Dinkelmeyer
Final year, UCL Economics with a Year Abroad
Winner - Third Best Paper
and Presentation
“One of the greatest tragedies of extreme poverty is its intergenerational transmission.
Children who grow up in poor families tend to be in poorer health and have lower levels
of education. They thus enter adulthood without “the basic capabilities” necessary to
take advantages of labour-market opportunities to pull themselves out of poverty and
enjoy an acceptable quality of life.” (Amartya Sen, 1999)
Discrimination is universally accepted as an unacceptable action. Whether it is discrimination based
on gender, ethnic minority or origin of birth all are intolerable actions that deny social participation or
recognition of human rights. Some forms of discrimination however receive more attention than
others. While all forms are unacceptable, with the following essay I would like to draw attention to
one kind of discrimination that could be described as hidden but nevertheless widespread in many
dimensions of civil life. Social origin is not immediately observable, yet evidence suggests that family
background is a strong predictor of future life chances, earnings, health and wealth. In this sense
inequality of opportunity could be interpreted as a form of discrimination based on family background
as a whole.
This essay attempts to break down to what extend the lottery of birth affects one’s life chances and
how economic advantages and disadvantages are transmitted across generations. My motivation for
this topic comes from my interest to develop a better understanding of how inequality of opportunity
arises and to what extend it may be reasonable to implement policy recommendations that support
social mobility.
In doing so this essay is divided into four sections. Section one will provide the reader a general
background of social mobility cross-different countries and provide motivation for the remaining
chapters. The following section will provide a review of the key literature dealing with the mechanism
of the transmission of economic advantages and disadvantages across generations. Section three will
introduce the methodology of the empirical research of this paper. The objective of this section is to
“break” Germany up into its sixteen Federal States and investigate whether or not social fluidity
differs across the different states. This study is based on the German Socio-economic panel provided
by the DWI (Deutsches Institut für Wirtschaftsforschung) - a longitudinal annual survey of
approximately 11,000 private households starting in 1984 containing a rich set of economic, social
and cultural variables. The final section will interpret results of this paper and reflect on other studies
in this context.
Motivation
Comparative studies on social mobility show a significant dispersion of the degree of social fluidly
across developed nations. Argentina and Peru, the United States and Canada or even Italy and Spain
are examples of countries where researchers report significantly different estimates of social mobility
despite a comparably similar settings in terms of politics, economics and geography.
92
In his speech at the Centre for American Progress Allen Kruger confirmed these findings and noted
the importance of research on social mobility. He introduced a chart from Corak (2012) that is now
commonly referred to as the “Gatsby curve”. It highlights the adverse relationship between social
mobility and the income distribution. The chart raised concern about connection of social mobility
and equality of income and what may cause these divergences. This puzzle provides motivation for
researchers to identify the drivers of social mobility.
Review of literature
This section aims to provide the reader a better understanding of how social advantages or
disadvantages are passed on across generations. While this is not an attempt to develop an exhaustive
literature review I aim to introduce three factors that are largely accepted by the academic community
to be key drivers of social mobility. Firstly a review of prominent literature on the impact of early
childhood development. Secondly a review of literature on parental influence on school choices. And
finally a review of literature of the influence of cultural capital within a family.
i.
Early Childhood development
“There is now substantial and consistent evidence that the family milieu during early childhood is
decisive for later achievement, such as educational attainment, earnings and careers, and also for
later social problems, such as dropping out of school and criminality. “ (Esping Andersen 2004).
Two prominent studies demonstrating the effect of early childhood development are the Perry
Preschool Program and the Abecedarian Program. Both studies attempt to assess the effects support of
a cohort of randomly selected disadvantaged children during early childhood. The Abecedarian
Project was a year round childcare intervention focusing on educational development from infancy
through the age of 5 years. The Perry Preschool Program was a 30 weeks preschool programme that
93
consisted of 2.5 hours training during weekdays and a 90 minutes home visit by a teacher during
weekends. Both programmes applied educational activities including games and focused on the
development of social, emotional and cognitive abilities.
Long-term follow-ups on the participants concluded significant effects into adulthood of both
programmes. “Children who participated in the early intervention program had higher cognitive test
scores from the toddler years to age 21. Academic achievement in both reading and math was higher
from the primary grades through young adulthood. Intervention children completed more years of
education and were more likely to attend a four-year college.” (www.abc.fpg.unc.edu)
Positive effects were documented even long after the intervention. The following figure summarises
some of the findings from the official report and concludes a positive long-term impact on social,
economic and criminal behaviour and educational attainment.
Source: http://www.highscope.org/content.asp?contentid=219
Empirical research also helped to better understand factors that contribute to a disadvantages in the
early childhood development. Knudsen, Heckman, Cameron and Shonkoff (2006) states that the most
extensively studied risk factors are poverty, limited parental education, parental mental health
problems, neglect or exposure to physical violence.
“There is also strong evidence that family instability, parental unemployment, and alcoholism
seriously impair children’s educational attainment” (Esping Andersen 2004).
According to Danzinger and Waldfogel (2000) in particular the period of growth for children between
birth and age of six are the most significant window for development of basic abilities in a lifetime.
ii.
School choices
Parental resources are likely to influence children’s educational choices and career planning.
“Education is a process that proceeds in stages, and early educational career decisions have a strong
effect on the choices available at later stages” (Dustman 2004)
The education systems in most developed countries are segregated in different stages, often primary
school, secondary school, and further or advanced education. At some of these stages, pupils face
94
decisions whether or not to study with a focus on vocational or formal academic training. These
decisions will impact their future earnings power and total duration of their education, which present
risks for their families and parents.
Christian Dustman (2004) studied the influence of parental background on school choices in
Germany. It is important to note that the Germany education system does not perform ability tests or
charge tuition fees on any form of education. Some may argue that this may support a particular
educationally mobile society. “In the absence of tuition fees for secondary schools (at least for recent
cohorts), and no selective entry tests, why do not all parents send their children to upper track
schools?”. Dustman (2004) argues that parents with weaker educational background may be less
confident about their children’s prospects and behave in a more risk adverse fashion when facing
decisions about the educational path of their children.
Information about the potential return on education, career planning, appreciation for education and a
sense for appropriateness differentiate households across the different position on the distribution of
the income curve. “In particular, we find that young adults who experience single parenthood as
children and those who come from families in the bottom income quartile have significantly lower
educational attainments, while those whose parents are homeowners, particularly outright owners,
have much higher attainments.” (Ermisch and Francesconi 2001). This leads children from more
affluent backgrounds to achieve higher levels of education and experience.
iii.
Quality of parenting and cultural capital within families
A growing literature on Cultural capital, as initially introduced by Bourdieu and Jean-Claude Passeron
(1973), focuses on the non-financial family assets as a source of advantage or disadvantage for
children born in developed countries. Unlike economic or social capital, cultural capital is composed
of a combination of knowledge, skills and education that contribute to a families’s status in society.
“Bourdieu states that cultural capital consists of familiarity with the dominant culture in society, and
especially
to
understand
and
use
educated
language."
Sullivan (2001)
Sociologists measure cultural capital by indicators that are linked to the consumption of cultural
resources such as literature, discussing cultural issues and others. Max Weber (1986) suggests similar
formations in which social groups differentiate themselves by prestige, honour and religion.
Families with higher cultural resources are able to provide their children with advantages in a range of
periods that are important in the formation of a career. This is present in different stages in a person’s
life and observable during the education, the process of entering the labour market and promotions
later in a person’s career.
Sullivan (2001) provides empirical evidence for the intergenerational transmission of cultural capital
from parents to children and suggests a significant effect on the performance in the GCSE (General
Certificate of Secondary Education) examinations. While the magnitude of the impact of cultural
capital on educational attainment is subject to debate other studies, including DiMaggio and Mohr
(1985), have confirmed a significant relationship. In their 1985 study the authors conclude significant
positive effects of cultural capital in the “educational attainment, college attendance, college
completion, graduate education, and marital selection”.
While different views exist in the definition of cultural capital and proxies that are applied to extract
information about cultural capital from surveys and datasets, both papers demonstrate the transition of
cultural capital from parents to their children and the impact on educational attainment and other areas
in a person’s career development.
95
On a similar note to household cultural capital, parents have a direct influence on their children’s
successful development through parenting and informal education. Children growing up in low
income families may be disadvantaged due to potentially lower quality of parenting and informal
education from parents and their environment. “Poor parents might provide inferior non-monetrary
benefits to their children if low income increases stress or makes it harder to obtain information
about the importance of non-monetary inputs” (Mayer and Leonard 2004). Low-income families are
more prone to fluctuations and shocks in their income, which can result in stress and lower resources
for the parenting and the development of their children. “Alternatively low income may cause parents
to develop patters of thought and behaviour such as lower expectations that are helpful in coping with
poverty but damaging to their children’s development.” (Mayer 1997 chapter 3)
Empirical research
The objective of this chapter is to find quantitative estimates for social mobility in Germany by
federal state in order to attempt a spatial analysis.
The empirical research of this paper is based on the study by Philipp Eisenhauer and Friedhelm
Pfeiffer (“Assessing intergenerational earnings persistence among German workers”, 2008). The
novelty of this analysis comes from a separate consideration of the sixteen federal States of Germany,
as opposed to treating them all as one. Sample adjustments were made in order to measure social
mobility by states. These are outlined below.
Data
My analysis is based on v29 of the G-SOEP (German Socio-economic Panel data) ranging from 1984
to 2012. This is a longitudinal study of private households in Germany. It surveys annually nearly
15,000 households and about 25,000 individuals about family composition, occupational biographies,
health, earnings and other indicators.
Real earnings are adjusted for real GDP-Growth to bring observations from different survey years on
a comparable basis.
Econometric approach
This study will estimate social mobility by a computation of intergenerational earnings elasticity. The
following regression is applied to a selected sample in the German Socio Economic Panel.
Ln Yi,t = α + β ln Yi,t-1 + ε
The β term will reveal information about the class movements, measured by income, in Germany.
This analysis will be conditioned on the different federal states to attempt a spatial analysis of social
mobility.
In the following estimation we will apply the following variables:
Log(fathers_income) = α + β log(son_income) + ε
Sample Summary
Groups excluded from the sample
Measures of Economic status
Age restriction
o Individuals that permanently
moved away from their region of
birth
o Migrants
Son
Five year average monthly earnings
Father
Five year average monthly earnings
Between 30 – 45
96
Notes on methodology
Its important to note why this study excludes Individuals that permanently moved away from their
region of birth from the estimation sample. Ultimately this study attempts to estimate social mobility
the federal states in Germany. As such it excludes individuals that have moved away from their place
of birth (potentially to other federal states) in order to focus on an accurate estimation of social
mobility that is attributable to the circumstances in the respective regions.
Dealing with sources of biases
i.
Measurement error in permanent income
This study will use an estimation of the permanent income as an approximation for economic status.
Permanent income is not observable for researchers and may be estimated by aggregating life-time
earnings of an individual and discount it to the net present value for each year that he spends in the
labour market. In our analysis we will rely on 5 years averages of earnings as a proxy for permanent
income. This reduces the volatility of our earnings estimates and simplification suggested by Solon
(1992).
ii.
Unrepresentative groups of subsample
First we exclude individuals that have permanently moved from the region (town level) of birth. In
order to compare social mobility in the different federal states it is necessary to eliminate potential
spill over effects between states.
Secondly migrant workers were removed from the sample population. Migrants are a not
representative subsample in our population and may distort the picture of social mobility.
Finally this study also restricts the age of individuals of interest between 30 and 45 years. This is to
ensure that we observe fathers and his offspring at a similar stage in their life and earning cycle.
Sample statistics
Statistic
Est. permanent income in
Euro
Sd. of est. permanent income
Age in years
Age – Difference in years
Number of observations
Findings
i.
Fathers
1301
597
44
Sons
1268
607
31
13
1054
Basic results
The regression of son on father log-earnings resulted in the following estimates. Our estimates of the
intergenerational earnings elasticity β=0.163 are statistically significant at a 95% confidence level.
97
iii.
Comparison with other studies
These findings have been confirmed by several other studies. While the estimate of this study is at the
lower range of the spectrum of different estimates we note that our sample selection may have
excluded a proportion of (highly mobile) high earners. This may offers an explanation of the lower
estimation.
Study
OLS Results
Couch and Dunn (1997)
0.124 (0.07)
Wiegand (1997)
0.238 (0.06)
Vogel (2007)
0.266 (0.06)
Eisenhauer and Pfeiffer (2008)
0.282 (0.09)
This study
0.1652 (0.04)
Source: Eisenhauer and Pfeiffer (2008), page 21
iv.
IV Results
0.402 (0.13)
0.374
-
Spatial analysis: Social mobility by federal state
The map below shows the results of the estimation of the intergenerational income elasticity by the
sixteen federal states. The spectrum of colors ranging from green to red represents a low
intergenerational elasticity (i.e. more socially mobile) in the green states and a higher level of elasticy
(i.e. more socially immobile) in the red states.
98
Conclusion
i.
Interpretation of spatial analysis
Estimating mearures for social mobility in the sixteen federal states of Germany has produced a
pattern that could be interpreted with economic reasoning. One may analyse this by grouping the
different federal states by commong economic characteristics and social mobility estimates. One
suggested formation may be as follows:
Group I (Bayern, Baden-Wuerttemberg, Hessen)
Group II (Hessen, Sachen-Anhalt, Sachsen)
Group III (Saarland, Rheinland –Pfalz, Nordrhein Westfalen)
Group IV (Berlin, Brandenburg, Schleswig Holstein)
99
Interpreting these results requires a deep understanding of the German economic system and in
particular how these states differ from one and another. This is beyond the scope of this study but I
hope to inspire other others by pointing out the arguably non-random nature of the distribution of
these patterns.
In an attempt to reflect on the literature review in the previous sections, in what follows I will apply
an instrument to the estimation results to investigate whether or not the research on early childhood
development is consistent with our findings.
a) Early Childhood development
In an attempt to analyse the impact of early childhood development I will instrument Kindergarten
participation as a proxy of child care on the intergenerational earnings elasticity by state.
The following chart shows intergenerational elasticity on the horizontal axis, the percentage of
children between 3-5 in childcare at local Kindergarten on the vertical axis and the population of the
federal states represented by the bubble size.
98
[CELLRANGE]
% Kindergarten particpiation 3-5 year olds
96
[CELLRANGE]
94
[CELLRANGE]
[CELLRANGE]
92
[CELLRANGE]
[CELLRANGE]
[CELLRANGE]
[CELLRANGE]
[CELLRANGE]
[CELLRANGE]
[CELLRANGE]
90
88
[CELLRANGE]
86
[CELLRANGE]
84
[CELLRANGE]
[CELLRANGE]
82
[CELLRANGE]
80
78
0.110
0.120
0.130
0.140
0.150
0.160
0.170
0.180
Intergenerational earnings elasticity
Source: Official report 2012 National office of Statistics. Page 36. Statistisches Bundesamt
Kindertagesbetreuung in Deutschland 2012, Begleitmaterial zur Pressekonferenz am 6. November
2012 in Berlin
As expected by the arguments of Esping Andersen (2004) and the studies on the Perry Preschool
Program and the Abecedarian Program reviewed in section III we observe an inverse relationship
between the level of childcare and social mobility. The dense population in the top left of the chart
represents states with relatively high child care efforts and intergenerational earnings elasticity. While
100
we cannot interfere a causal relationship we can confirm that higher social mobility tends to occur in
areas with a stronger childcare.
ii.
Conclusion
This study has shown that social mobility varies both between and within countries. Estimating
mearures for social mobility in the sixteen federal states of Germany has produced a pattern that
could be interpreted with economic reasoning.
Bibliography
Sen, Amartya 1999. Development as freedom. Oxford, U.K.: Oxford University Press
Alice Sullivan, Cultural Capital and Educational Attainment Sociology November 2001 35: 893-912,
Andreas Björklund, Tor Eriksson, Markus Jäntti, Oddbjorn Raaum and Eva Österbacka "Family structure
and labour market success: the influence of siblings and birth order on the earnings of young adults in
Norway, Finland and Sweden” Chapter 9 in Corak M, ed. Generational Income Mobility in North
America and Europe. Cambridge: Cambridge University Press
Becker G. S., Tomes N. (1979). An Equilibrium Theory of the Distribution of Income and
intergenerational Mobility. Journal of Political Economy vol. 87 no. 6j.
Becker G.S. (1964). Human Capital: A Theoretical and Empirical Analysis, with Special Reference to
Education, 2nd ed. New York
Bourdieu and Passeron 1973. Cultural Reproduction and Social Reproduction. In Richard K. Brown
(Ed.), Knowledge, Education and Cultural Change. London: Tavistock
Christine R Schwartz Robert Mare .Trends in Educational Assortative Marriage From 1940 to 2003.
Demography, Volume 42, Number 4, November 2005, pp. 621-646 (Article)
Corak, Miles (2004). “Introduction.” In Miles Corak (editor). Generational Income Mobility in North
America and Europe. Cambridge: Cambridge University Press.
Corak, Miles (2004). “Introduction.” In Miles Corak (editor). Generational Income Mobility in North
America and Europe. Cambridge: Cambridge University Press.
Corak, Miles. 2004. "Generational income mobility in North America and Europe: an introduction.
Chapter 1 in Corak M, ed. Generational Income Mobility in North America and Europe. Cambridge:
Cambridge University Press.
Eric Knudsen & James J. Heckman & Judy Cameron & Jack P. Shonkoff, 2006. "Economic,
Neurobiological and Behavioral Perspectives on Building America’s Future Workforce," World
Economics, World Economics, Economic & Financial Publishing, PO Box 69, Henley-on-Thames,
Oxfordshire, United Kingdom, RG9 1GB, vol. 7(3), pages 17-41, July.
Erikson and Goldthorpe. 1992. The Constant Flux: A Study of Class Mobility in Industrial Societies.
Oxford: Clarendon Press.
GORDON MARSHALL. "Goldthorpe class scheme." A Dictionary of Sociology. 1998. Retrieved
December
06,
2014
from
Encyclopedia.com:
http://www.encyclopedia.com/doc/1O88-
101
Goldthorpeclassscheme.html
Gøsta Esping-Andersen 2004. "Unequal opportunities and he mecahnisms of social inheritance.” Chapter
12 in Corak M, ed. Generational Income Mobility in North America and Europe. Cambridge: Cambridge
University Press
Gøsta Esping-Andersen 2004. "Unequal opportunities and he mechanisms of social inheritance.” Chapter
12 in Corak M, ed. Generational Income Mobility in North America and Europe. Cambridge: Cambridge
University Press
Jacob A. Mincer, 1974. "Summary and Agenda," NBER Chapters, in: Schooling, Experience, and
Earnings, pages 128-144 National Bureau of Economic Research, Inc.
Jo Blanden, Alissa Goodman, Paul Greeg, and Stepen Machin "Changes in the intergenerational mobility
in Britain.” Chapter 5 in Corak M, ed. Generational Income Mobility in North America and Europe.
Cambridge: Cambridge University Press
John Ermisch and Marco Francesconi .Family Matters: Impacts of Family Background on Educational
Attainments. Economica, New Series, Vol. 68, No. 270 (May, 2001), pp. 137-156
Kerwin Kofi Charles, Erik Hurst, Alexandra Killewald. Marital Sorting and Parental Wealth .
Demography February 2013, Volume 50, Issue 1, pp 51-70
Lemieux, Thomas. (2006) "The 'Mincer equation' Thirty Years after Schooling, Experience, and
Earnings" in Jacob Mincer: A Pioneer of Modern Labor Economics, Shoshanna Grossbard, ed., Springer:
New York. pp. 127–145
Mayer Susan E., Lopoo Leonard M. "What do trends in the intergenerational economic mobility of sons
and dauthers in the United States mean?.” Chapter 5 in Corak M, ed. Generational Income Mobility in
North America and Europe. Cambridge: Cambridge University Press
Mayer Susan E., Lopoo Leonard M. "What do trends in the intergenerational economic mobility of sons
and dauthers in the United States mean?.” Chapter 5 in Corak M, ed. Generational Income Mobility in
North America and Europe. Cambridge: Cambridge University Press
Mazumder, Bhashkar (2005). “Fortunate Sons: New Estimates of Intergenerational Mobility in the
United States Using Social Security Earnings Data.” Review of Economics and Statistics. Vol. 87, No. 2
(May), pages 235-255.
Miles Corak 2012. “Inequality from generation to generation: the United States in comparision”. (2012).
In Robert Rycroft (editor), the economics of inequality, poverty and discrimination in the 21st century.
ABC-CLIO. In press, publication date February 2013.
Miles Corak and Patrizio Piraino. The Intergenerational Transmission of Employers. Journal of Labor
Economics, Vol. 29, No. 1 (January 2011), pp. 37-68
Paul DiMaggio and John Mohr ,Cultural Capital, Educational Attainment, and Marital Selection
American Journal of Sociology Vol. 90, No. 6 (May, 1985) , pp. 1231-1261 Published by: The
University of Chicago Press Stable URL: http://www.jstor.org/stable/2779635
102
Rames (2000). Ramses H. ABUL NAGA, 2000. "Galtonian Regression of Intergenerational Income
Linkages : Biased Procedures, a New Estimator and Mean-Square Error Comparisons," Cahiers de
Recherches Economiques du Département d'Econométrie et d'Economie politique (DEEP) 00.13,
Université de Lausanne, Faculté des HEC, DEEP.
Schnitzlein, Daniel D. (2013): Low Level of Equal Opportunities in Germany: Family Background
Shapes Individual Economic Success. DIW Economic Bulletin. 3 (5), 3-8
Solon 1991 G. Solon, M. Corcoran, R. Gordon, and D. Laren, »A Longitudinal Analysis of Sibling
Correlations in Economic Status,« The Journal of Human Resources 26 (3) (1991): 509–534.
Solon, G. (2002). Cross-Country differences in intergenerational earnings mobility. Journal of Economic
Perspectives16> 59-66
Solon, Gary, 1999. "Intergenerational mobility in the labor market," Handbook of Labor Economics, in:
O. Ashenfelter & D. Card (ed.), Handbook of Labor Economics, edition 1, volume 3, chapter 29, pages
1761-1800 Elsevier.
Susan Mayer What Money Can't Buy: Family Income and Children's Life Chances. (1997), Cambridge
MA: Harvard University Press.
103
The effects of television viewing on income aspirations and
happiness by Emma Claydon
Final Year, UCL Economics with a Year Abroad
Happiness economics is regaining attention in modern economic and political research; a focus that
hasn’t been as strong since the 18th century with the rise of utilitarianism. Foremost at that time,
Adam Smith (1790) argued that ‘All constitutions of government, however, are valued only in
proportion as they tend to promote the happiness of those who live under them. This is their sole use
and end.’ More recently the debate regarding happiness economics has explored the effect of relative
income and governments have also sought to better understand the relationship between traditional
measures of economic well-being and the happiness of their citizens. In this essay I explore the
relationship between income and happiness. Subsequently, I examine the effects of television
viewing, as a global everyday activity that could be argued to have a major influence on societal
ideals, on both income aspirations and happiness. To do this, this paper first reviews existing literature
surrounding the topic, and then contributes to this with further empirical analysis.
The income-happiness paradox.
More recent debates regarding happiness economics have stemmed from Easterlin’s (1974) seminal
paper, in which he found that growth in the per capita income of a country does not lead to increased
levels of human happiness. Subsequently, the ‘happiness paradox’ has also become known as the
‘Easterlin paradox’. From 30 surveys during a 25 year period in the USA, Easterlin found that per
capita real income rose by over 60 per cent. The proportions of people, however, who rated
themselves as ‘very happy’, ‘fairly happy’, or ‘not very happy’ stayed almost unchanged. The
majority of similar studies also don’t find an increase in happiness over time, despite the fact that the
per capita real income of each country may be increasing dramatically. As R. Skidelsky and E.
Skidelsky (2012) argue, this suggests that happiness is affected by relative income, rather than
absolute income; since the rich remain at the top end of the distribution, and the poor at the bottom,
average happiness levels do not change with the income of the country as a whole.
The relative income hypothesis was first introduced by Duesenberry (1949) in ‘an attempt to
rationalize the well established differences between cross-sectional and time-series properties of
consumption data’ (Alvarez-Cuadrado and Van Long, 2011). The relative income hypothesis states
that the utility, or satisfaction, that one gets from a certain level of consumption depends on how it
sizes up relative to that of others, rather than in an absolute manner. Alternatively, the absolute
income hypothesis states that utility should always increase with the level of income, although not
necessarily at the same rate.
Explanations for, and mechanisms behind, the income-happiness paradox.
There are many explanations for the Easterlin paradox, the major ones of which may be divided into
the following categories: positional, hedonic and satisfaction explanations (Bruni and Stanca, 2006).
The most relevant to this paper, and perhaps the most important amongst economists, is the positional
treadmill hypothesis. Hirsch (1977) coined the term ‘positional good’, referring it to a good whose
consumption, and utility thereof, depends negatively on the consumption of the same good by others.
In Solnick and Hemenway’s (1998) study, they asked members of faculty, staff and students at
Harvard School of Public Health to choose between two possible states of the world; one in which
104
their annual income is $50,000 and others earn $25,000, or a state in which they would earn $100,000
and others would earn $200,000. The first state was the ‘positional case’, where the participant has
more than others in society. The latter state was the ‘absolute case’, where both the participant and the
others would earn a greater absolute amount than in the positional case, but the participant would
have less than others in society. They found that approximately 50% of the participants would prefer a
world in which they had half as much real income, as long as it was a relatively high income. This
demonstrates the importance of positional concerns and taking account of relative income, as well as
absolute.
Frank (1991) argued that positional externalities occur when `one person's action alters an important
frame of reference for others' and that the ‘positional treadmill’ occurs when all are trying to get an
advantage over each other. This is because there is no gains if everyone makes these shifts as their
relative positions stay the same (Frank, 1985). Changes in the characteristics of the reference group,
however, could alter one’s frame of reference. For example, Layard (2005b) recognised from
Oswald’s prior research that ‘a rise in the average income in the state where you live reduces your
happiness by one-third as much as a rise in your own income increases it’. He further recognised that
‘a rise in the wages of comparable workers reduces your job satisfaction by as much as a rise in your
own wage increases it’. This again demonstrates both the sensitivity to, and importance of, relative
income, and how this can lead to rising material and income aspirations.
The effects of material aspirations on happiness.
Increasing material aspirations has been argued to be one of the main factors behind the incomehappiness paradox. This is due to the fact that an individual’s subjective well-being depends on the
distance between one’s actual material achievements and one’s aspirations, rather than their actual
income (Stutzer, 2004). Following this logic, if an individual’s income aspirations are to increase then
their utility will fall, unless their actual income increases by the same amount or greater. Stutzer
(2004) also finds that income aspirations increase with the average income in the community in which
they live, as well as when their own income increases. This is consistent with the theories of
habituation and positional concerns. In particular, Stutzer found that the estimated effects on one’s
aspirations of a higher average community income are greater for people who actively interact more
with other members of the community. This adheres to the finding that increases in reference values
are found to exert a negative effect on one’s happiness, even after controlling for absolute income and
other personal or demographic factors (Caporale et al., 2009).
Over time, happiness may not change as income rises, should both aspirations and income rise
together. It is, therefore, important to look at what may affect material aspirations. I will explore
television viewing as a major determinant of income aspirations, namely through the channels of
increasing and widening social comparisons (positional concerns) and through speeding up the
satisfaction treadmill.
Television viewing as a determinant of material aspirations.
Watching television plays a major part in most people’s lives, with the average TV viewer in the UK
watching 235 minutes of television each day (Thinkbox, 2013). The significance of this figure is
evident when considered against the fact that the average Briton has 998 waking minutes per day
(Von Radowitz, 2012), meaning that they are spending almost a quarter of their waking life watching
television. It is, therefore, very plausible to suppose that an activity that has such a large quantity of
time dedicated to it will bear some effect on individuals’ material aspirations and, consequently, on
their well-being.
105
According to cultivation theory, ‘the more time people spend 'living' in the television world, the more
likely they are to believe social reality portrayed on television’ (Cohen and Weimann, 2000), and
‘Heavier viewers tend to believe luxury products and services to be more commonplace than they
actually are’ (Shrum, Burroughs & Rindfleisch, 2005). Psychological effects, such as these, that
surround the issue of television viewing are well documented; the economic impacts, however, less
so.
Bruni and Stanca (2006) argue that watching television speeds up the satisfaction treadmill through
faster growth of aspirations. It can also speed up the positional treadmill through its effect on the
reference groups that one uses to form their aspirations. Television can widen these reference groups,
often with a skew towards increasing the number of wealthy within such a group due to the fact that
television shows ‘many more affluent people and more luxury than exist in real life’ (Frey, Benesch
& Stutzer, 2007).
Hyll and Schneider (2013) explored whether TV viewing has an impact on material aspirations by
using a natural experiment that took place when Germany was divided between West and East. There
was a natural variation in exposure to certain types of television, with certain regions of East
Germany receiving little or no advertising and with material preferences depicted negatively. Some
regions of East Germany, however, could receive both East and West German TV channels, with the
Western ones being consumption orientated, influenced by the US, UK and France and emphasising
materialistic values and individualistic lifestyles. They found that watching West German television
had a positive and highly significant effect on income aspirations, consumption aspirations and
hedonistic aspirations. In particular, it was found that, on average, if one extended their West German
television consumption from ‘never’ to ‘daily’, their probability of expressing very high consumption
aspirations would rise from 9.6% to 14.1%, implying a causal relationship. This further enforces
claims that income and consumption comparisons are made against characters, messages and images
seen on television, as well as against friends and the local community.
Television, income aspirations and happiness.
There are a limited number of papers that aim to link all of these steps together in terms of economic
impacts; the effect of watching television on material aspirations and, in turn, the effect of one’s
income on their happiness. The primary two papers to address all of the above have been produced by
Frey et al. (2007) and Bruni and Stanca (2006). Frey et al. explore television viewing as a situation
where the theory of revealed preference does not hold. They argue that many individuals may overconsume TV due to a lack of self-control or foresight into their own behaviour and that this is not
utility maximising. This can be partly due to the benefits of television consumption often being shortterm and immediate, whereas there are many long-term costs that can be hard to predict and can take
a lot longer to appear; for example, ‘an increase in one’s material aspirations might not be foreseen at
all’. Bruni and Stanca (2006) provide a detailed analysis of the mechanisms behind material
aspirations, arguing that ‘television has a powerful effect on the satisfaction an individual derives
from his income and consumption levels by speeding up both the hedonic and positional treadmills.’
They find that, whilst income is positively correlated with life satisfaction, this positive effect is
significantly smaller for those who watch a high amount of television. Furthermore, this effect is
magnified when looking at financial satisfaction in particular, rather than life satisfaction in general.
Bruni and Stanca reason that these results can be seen as an additional explanation for the incomehappiness paradox, in that as television viewing increases with living standards, material aspirations
are increased. This then lessens the effect of income on one’s happiness.
106
The significance of this study.
In this paper I intend to further explore the effects of watching television on life satisfaction and
happiness. I propose to subsequently investigate whether any discovered effect differs according to
the proportion of television consumption spent watching news programmes, as opposed to general
television programmes.
For the main part, my general hypothesis is to examine television viewing as a determinant of
individual aspirations. I will follow the belief that happiness is not affected solely by absolute income,
but by the gap between income and material, or income, aspirations. I will, therefore, argue that
television consumption increases material aspirations, resulting in reduced income satisfaction.
‘A house may be large or small; as long as the neighboring houses are likewise small, it satisfies
all social requirement for a residence. But let there arise next to the little house a palace, and the
little house shrinks to a hut.’ – Karl Marx (1849).
‘The generalised obsession with fame and wealth, the pervasive sense, in watching it [television],
that life is somewhere other than where you are.’ - George Monbiot (2014).
Data and methodology.
The data that I will use is from the 5th Round (2010) of the European Social Survey (ESS) [1], and
from this I will use the data of participants in the United Kingdom (UK). The ESS is conducted every
two years and has been in operation since 2001, in over 30 nations. It focusses on monitoring social
change by measuring the attitudes, beliefs, behavioural patterns and values within these nations. The
ESS dataset for the UK that this paper utilises consists of 1865 individuals, after omitting individuals
who did not answer the relevant questions. The data includes results regarding television and media
consumption, subjective rating of happiness and life satisfaction, as well as information on a number
of socio-economic factors that can be used as control variables.
Using the ESS data, I intend to first estimate the following regression:
SATLIFEi = β0 + β1 TVTOTi + β2 Xi + εi
Where ‘SATLIFE’, the dependent variable, is measured by the question of ‘All things considered,
how satisfied are you with your life as a whole nowadays?’, and acts as a proxy for individual
happiness. The answers to this are based on a scale of 0-10, with 0 being ‘Extremely dissatisfied’ and
10 being ‘Extremely satisfied’. ‘TVTOT’ represents the independent variable measured by the
question ‘On an average weekday, how much time, in total, do you spend watching television?’, and
is measured on a scale of 0-7, with 0 being ‘No time at all’ and 7 being ‘More than 3 hours’. Finally,
‘X’ represents a range of socio-economic control variables, including household income, gender, age,
subjective health, extent of religiousness, their marital status, and their main activity over the last
seven days e.g. paid work/unemployed/retired.
Figure 1 shows the distribution of the dependent variable data, regarding how satisfied one is with
their life. It shows that the majority of people are more satisfied than dissatisfied; however, there is a
fair spread which will provide enough variation for further analysis.
107
Figure 1: 'How satisfied are you with your life as a whole?'
700
No. of indivduals
600
500
400
300
200
100
0
Figure 2 displays the distribution of how much time people spend watching television in general
(TVTOT) and political/news programmes on television (TVPOL). The TVPOL data consists of
answers to the question ‘On an average weekday, how much of your time watching television is spent
watching news or programmes about politics and current affairs?’, and is answered using the same
measurement scale as TVTOT. Fig. 2, therefore, demonstrates that the most chosen answer for
TVTOT was ‘more than 3 hours’. It also shows that, for TVTOT, the data is clustered around the
higher end of the scale, whereas, for TVPOL it is the opposite case.
No. of individuals
Figure 2: 'On an average weekday, how much time, in total, do
you spend watching television?'
700
600
500
400
300
200
100
0
None at all
< 0.5 hr
0.5 hr to 1 hr
> 1 hr, up to 1.5 hrs
> 1.5 hrs, up to 2 hrs
> 2 hrs, up to 2.5 hrs
TVTOT
TVPOL
> 2.5 hrs, up to 3 hrs
This relationship is further explored in Figure 3, in which I have calculated a new variable; ‘tvprop’,
where tvprop = 1 – (TVPOL/TVTOT). Although the values of ‘tvprop’ aren't interpretable directly, it
does gives an idea of the distribution of the amount of time TV viewers spend watching general TV
programmes, as opposed to the news/political programmes. In this respect, it shows that the majority
of people in the dataset spend more time watching general TV than they do political/news TV
programmes. This could aid my analysis further in the paper as, according to my hypothesis, watching
general TV programmes should have a more detrimental effect on one’s happiness, as measured by
life satisfaction, than watching political programmes, such as the news. This is because general TV
108
programmes are more likely to extend the top, wealthy end of viewers’ reference groups. On the other
hand, political programmes are more likely to not affect one’s income aspirations and reference group
as much, or may even extend it at the bottom, poorer end, by making viewers more aware of those
worse off than them.
Figure 3: Proportion of time spent watching tv spent
on general (ie. non-news/political) programmes
No. of individuals
350
300
250
200
150
100
50
0
Proportion, with 1 = All general tv, no political/news
Results.
I first ran an OLS regression of SATLIFE on TVTOT (Appendix A). However, SATLIFE, the
dependent variable is a Likert-type item, measured on an ordinal scale. As a result, to use OLS one
would have to assume that the underlying concept is continuous, or that it is an interval level scale. In
this case, we cannot be sure that the individuals answering the life satisfaction question would, for
example, view the difference between ratings ‘1’ and ‘2’ as equal to the difference between ratings ‘5’
and ‘6’. Therefore, I decided to run an ordinal logistic regression of SATLIFE on TVTOT, which may
be more suited to such data Following this, I also ran the same ordinal logit regression, but omitting
the subjective health control variable, as this might have been absorbing large parts of the variation in
happiness. After omitting health, results were found for TVTOT that are statistically significant at the
10% significance level. Figure 4 shows a comparison of the results for the ordinal logit regressions
with, and without, the health control variable.
Firstly, to assess the overall empirical specification, I will look at some of the coefficients of the control
variables, as shown in Figure 4. At a 1% level of statistical significance, it is shown that the worse one regards
their health, the less satisfied one will be with their life as a whole. At this level it can also be seen that being
separated, divorced or having never been married has a negative impact on life satisfaction. Being unemployed,
but looking for work, has a negative effect on life satisfaction, whereas being retired has a positive effect of a
similar extent, both of which are significant at the 1% level.
Looking at TVTOT, from the regression that omits the health variable, it can be seen that watching between 1.5
and 2.5 hours of television, or more than 3 hours per day has a significant negative impact on life satisfaction, to
a 10% level of statistical significance. This supports the hypothesis that television has a negative effect on life
satisfaction, or happiness, and that increased income aspirations could be an explanation behind this.
However, it is possible that the issue of reverse causation could be present. In other words, it might be the case
that, rather than high levels of TV consumption causing a fall in happiness, people who are less happy may
watch more TV as a result of their unhappiness. Therefore, I will further explore each stage of my hypothesis to
shed some more light on the possible relationships present. Firstly, I am going to explore the difference in
109
effects of watching general television programmes (TVTOT), against those of watching the news and other such
political programmes (TVPOL).
An ordinal logit regression of SATLIFE on TVPOL (Appendix B) doesn’t find any significant results, even
when the subjective health variable is omitted, and also does not show such a strong negative trend as was
shown with TVTOT. This could infer the effect of television viewing on happiness as being through the
increased income and material aspirations mechanism, as a result of higher reference groups to compare oneself
to. This is due to the fact that one is more likely to be subjected to such wealth and luxury items in general TV
programmes rather than political ones. Further to this, the variable ‘tvprop’ takes a value between 0 and 1, and
the closer it is to 1, the more general television the individual watches as opposed to political TV. In an ordinal
logit regression of SATLIFE on tvprop (Appendix C), the coefficient of tvprop indicates a negative relationship
between the proportion of general TV watched to political, and life satisfaction, albeit not statistically
significant. This further supports the above hypothesis.
Figure 4.
110
Further analysis and results.
To attempt to confirm the mechanism through which television consumption affects happiness, I am
going to break down the hypothesis by looking separately at the first, most major stage; that TV
viewing increases income aspirations. The variable ‘IMPRICH’ consists of answers to the question of
how much the individual thinks the following statement applies to them; 'It is important to her/him to
be rich. She/he wants to have a lot of money and expensive things.' It is answered on a scale of 1 to 6,
where 1 essentially signifies that they find being rich to be very important, and 6 signifying that they
don’t find being rich to be important. This data is displayed in Figure 5. An ordinal logit regression of
IMPRICH on TVTOT (Appendix D) finds that, at the 5% significance level, watching either up to ½
an hour, or between 2 and 3 hours of TV per day is positively correlated with how important the
individual finds being rich and owning expensive things. This supports the major first stage of this
paper’s hypothesis; that television consumption increases material aspirations. Further to this, an
ordinal logit regression of IMPRICH on TVPOL (Appendix E) finds no statistically significant
correlations and there doesn’t appear to be any strong positive or negative trend. This infers again that
the effect of TV consumption on material/income aspirations may be through the mechanism of being
subjected to a more wealthy reference group and more luxury items.
The above infers that first stage of the hypothesis holds; that television consumption increases
income, or material, aspirations. I will now separately examine the second stage of the hypothesis;
that being less satisfied with one’s income, which can be caused by increased income aspirations,
causes a fall in happiness, or life satisfaction as a whole. An ordinal logit regression of SATLIFE on
HINCFEEL (Appendix F) shows that the worse one feels about one’s income, the lower their overall
life satisfaction will be, and this is significant to a 1% level. The variable ‘HINCFEEL’ is shown in
Figure 6, and is the responses to the question 'Which of the descriptions comes closest to how you feel
about your household's income nowadays?'
Figure 5: 'It is important to her/him to be rich. She/he wants
to have a lot of money and expensive things.'
1200
No. of indivduals
1000
800
600
400
200
0
1 - Very
much like
me
2 - Like me
3Somewhat
like me
4 - A little
like me
111
5 - Not like 6 - Not like
me
me at all
Figure 6: 'Which of the descriptions comes closest to
how you feel about your household's income
nowadays?'
5%
1 - Living comfortably on
present income
14%
36%
2 - Coping on present
income
3 - Difficult on present
income
4 - Very difficult on
present income
45%
Conclusion.
This paper finds that watching between 1.5 and 2.5 hours of television, or more than 3 hours per day,
has a significant negative impact on life satisfaction, to a 10% level of statistical significance. This
supports the hypothesis that television consumption can have a negative effect on life satisfaction, or
happiness. Increased income aspirations resulting from being subjected to a more wealthy reference
group whilst watching television programmes could be a major explanation behind this. In favour of
this mechanism, this paper finds that watching political programmes, such as the news, does not have
such an effect on life satisfaction, possibly due to the fact that these programmes do not purvey such
levels of wealth, consumerism and luxury items.
Further to this, in order to fully comprehend how the relationship between TV viewing and happiness
may function, this paper finds that watching television can be positively correlated with how
important the individual finds being rich and owning expensive things. The importance that an
individual places on this will directly shape their material and income aspirations, and so this result
supports the hypothesis that TV consumption can lead to increased income aspirations. This paper
further confirms that the worse one feels about one’s income, the lower their overall life satisfaction
will be, thereby completing support for each stage of the hypothesis that television viewing can cause
lower happiness through the mechanism of increased income aspirations.
Bibliography
2013 TV Viewing. (n.d.). Retrieved 10 January 2015, from http://www.thinkbox.tv/tv-at-a-glance/2013tv-viewing/
Alvarez-Cuadrado, F., & Van Long, N. (2011). The relative income hypothesis. Journal of Economic
Dynamics and Control, 35(9), 1489–1501.
Bruni, L., & Stanca, L. (2006). Income Aspirations, Television and Happiness: Evidence from the
World Values Survey. Kyklos, 59(2), 209–225.
112
Bruni, L., & Stanca, L. (2008). Watching alone: Relational goods, television and happiness. Journal
of Economic Behavior & Organization, 65(3-4), 506–528.
Caporale, G. M., Georgellis, Y., Tsitsianis, N., & Yin, Y. P. (2009). Income and happiness across
Europe: Do reference values matter? Journal of Economic Psychology, 30(1), 42–51.
Cohen, J., & Weimann, G. (2000). Cultivation revisited: Some genres have some effects on some
viewers. Communication Reports, 13(2), 99–114.
Di Tella, R., & MacCulloch, R. (2006). Some Uses of Happiness Data in Economics. Journal of
Economic Perspectives, 20(1), 25–46.
Duesenberry, J. (1949). Income, Saving and the Theory of Consumer Behaviour. United States:
Harvard University Press.
Easterlin, R. (1974). Does Economic Growth Improve the Human Lot? Some Empirical
Evidence. Nations and Households in Economic Growth, 89–125.
Frank, R. (1987). Choosing the Right Pond: Human Behaviour and the Quest for Status (1st ed.).
United States: Oxford University Press Inc.
Frank, R. H. (1991). Positional Externalities. Strategy and Choice: Essays in Honor of Thomas C.
Schelling, 25–47.
Frey, B., Benesch, C., & Stutzer, A. (2007). Does watching TV make us happy? Journal of Economic
Psychology, 28(3), 283–313.
Hirsch, F. (1976). Social limits to growth. United States: Cambridge, Mass.β€―: Harvard University
Press, 1976.
Holt-Lunstad, J., Smith, T., & Layton, B. (2010). Social Relationships and Mortality Risk: A Metaanalytic Review. PLoS Medicine, 7(7).
Hyll, W., & Schneider, L. (2013). The causal effect of watching TV on material aspirations: Evidence
from the ‘valley of the innocent’.Journal of Economic Behavior & Organization, 86, 37–51.
Johns, H. and Ormerod, P. (2007) Happiness, economics and public policy. United Kingdom:
Londonβ€―: Institute of Economic Affairs, 2007.
Layard, R. (2005a). Happiness Lessons From The New Science. United States: Penguin Press HC,
The.
Layard, R. (2005b). Rethinking Public Economics: The Implications of Rivalry and Habit. Economics
and Happiness, 147–169.
Marx, K. (1975). Wage-labour and capital & Value, price, and profit (1st ed., p. Chapter 6). United
States: International Publishers Company, Incorporated.
Monbiot, G. (2014, October 14). The age of loneliness is killing us. The Guardian. The Guardian.
Retrieved from http://www.theguardian.com/commentisfree/2014/oct/14/age-of-loneliness-killing-us
Shrum, L. J., Burroughs, J. E., & Rindfleisch, A. (2005). Television’s Cultivation of Material
Values. Journal of Consumer Research, 32(3), 473–479.
Skidelsky, E., & Skidelsky, R. (2012). How Much Is Enough?: Money and the Good Life. United
Kingdom: Other Press (NY).
Smith, A. (1790). The Theory of Moral Sentiments. Library of Economics and Liberty. Retrieved from
January 30, 2015 from the World Wide Web: http://www.econlib.org/library/Smith/smMS4.html
113
Solnick, S., & Hemenway, D. (1998). Is more always better?: A survey on positional
concerns. Journal of Economic Behavior & Organization, 37(3), 373–383.
Stutzer, A. (2004). The role of income aspirations in individual happiness. Journal of Economic
Behavior & Organization, 54(1), 89–109.
Stutzer, A., & Frey, B. S. (2010). Recent advances in the economics of individual subjective wellbeing. Discussion Paper Series // Forschungsinstitut Zur Zukunft Der Arbeit, No. 4850.
Von Radowitz, J. (2012, June 12). British adults sleep more than seven hours a night on average. The
Independent. Independent. Retrieved from http://www.independent.co.uk/news/uk/home-news/britishadults-sleep-more-than-seven-hours-a-night-on-average-7843643.html
[1] ESS Round 5: European Social Survey Round 5 Data (2010). Data file edition 3.2. Norwegian
Social Science Data Services, Norway – Data Archive and distributor of ESS data.
ESS Round 5: European Social Survey (2014): ESS-5 2010 Documentation Report. Edition 3.2.
Bergen, European Social Survey Data Archive, Norwegian Social Science Data Services.
114
Bringing the Vanguard to the Rearguard: A transformative
development model for Brazil’s backlands By James B Theuerkauf *
Final Year, UCL Economics
“This trajectory is about the progressive reconstruction of the institutional content of the market
economy, giving practical form to the ideal of economic freedom, and creating a world in which
prosperity and freedom are on the same side.”44 – Roberto Mangabeira Unger
1. Introduction
“E você, porque está aquí?!” (“And you, why are you here?!”). This was the question I was most
frequently asked while conducting the fieldwork research in the heart of Brazil’s poorest region: the
backlands. The region regularly suffers from severe droughts, resulting in very tough living
conditions for the population. GDP per capita and education levels are among the lowest in Brazil.
Throughout its history, the region has been neglected by policymakers and under-researched by
academics. As this research will show, the backlands have huge potential. However, this potential still
needs to be partially unlocked. The current development framework has not succeeded to do so, thus
disempowering the backlands population.
What the backlands need is a transformative development model that gives means and opportunities
to its people, so they can achieve economic empowerment. Such a model needs to take the region’s
reality today, which is characterised by significant socio-economic shifts, as its point of departure, to
then build upon it and propose a path for real, transformative development. Therefore, my focus has
been on first gaining an in-depth understanding of the actual reality of the backlands, in order to then,
based on this, propose and imagine an adjacently possible solution.
After conducting primary research in the backlands, and in conjunction with this thesis’ academic
advisor, Professor Roberto Mangabeira Unger, I argue that post-Fordism – a socio-economic theory
that describes vanguard forms of production – should be the centrepiece of such a model. I find that
many post-Fordist attributes are already present in the backlands today and dominate its regional
economy, so that a development model geared around it is both adequate and feasible. That said, I
also find that many fundamental elements are missing, posing a serious threat to sustainable
development. In order to address these problems and overcome them, I outline and argue for a set of
policy implementations, with the aim of providing functional equivalents to the missing attributes.
This essay is structured as follows: I will firstly illustrate the backlands’ economic reality, as well as
the current approaches to development, which are centred around poverty alleviation and large
industrial projects. Thereafter, I introduce a model of post-Fordism, which describes vanguard forms
of production on an individual worker level, a firm level, and an inter-firm level. I derive the model
through a recombination of the literature. This research project’s central question is if, and how, postFordism (the vanguard) and the backlands (the rearguard) can be united. Before providing my answer,
* B.Sc. Economics Candidate (L100) | University College London (Department of Economics) | Class of 2015. My undergraduate thesis in
economics has the same title as this essay, and will be submitted to the UCL Department of Economics in March 2015 in partial fulfillment
for the requirements of a Bachelor of Science in Economics. I would like to thank, in particular, Dr. Cloda Jenkins (UCL) for reviewing the
essay for this conference and providing excellent feedback and advice. Further, I would like to thank Professor Roberto Mangabeira Unger
(Harvard University) and Dr. Frank Witte (UCL) for outstanding mentorship, academic guidance, as well as constant encouragement during
the whole thesis writing process.
44
As presented in the “Visions For The Future” (2012) documentary series.
116
I briefly outline the method of primary research/fieldwork that I employed to derive my results.
Subsequently, I present, explain, and discuss my results: they show that some of the pivotal postFordist attributes are already present in the region, while others are still in transition or are outright
absent. From the results I conclude that post-Fordism is not only adequate as a development model for
the backlands, but that it is also viable. However, the absence of central attributes is a significant
problem that needs to be overcome. To conclude my thesis, I outline the public policies which could
provide functional equivalents in order to bring the vanguard to the rearguard.
“Espero contribuir a um modelo de desenvolvimento transformador para esta região” (“I hope to
contribute to a transformative development model for this region”). That was the one-line answer I
gave to most people asking the question that introduced this paper. The more elaborate answer – the
one that describes the problem, explains and discusses the work and results, and then proposes
solutions – is the topic of this essay.
2. The backlands’ socio-economic reality: past and present
Figure 1: Brazil’s Northeast
Figure 2: Brazil’s backlands
Source: Raphael Lorenzeto de Abreu
Source: Agência Nacional de Águes (ANA)
Located in the country’s Northeast (Fig. 1), the Brazilian backlands (Fig. 2) are home to 20.6 million
people, with a GDP per capita of R$7,158, which is only about a fifth of Brazil’s GDP per capita of
R$34,315. 45 Historically, the region has been associated with the occurrence of severe draughts
(Gomes, 2001; Araújo and Lima, 2009): as well as making the living conditions very difficult, the
droughts wiped out most of the region’s agriculture-based economy, with GDP levels falling by as
much as 75% 46 and hundreds of deaths occurring due to the lack of food and water (Khan and
Campos, 1995). In fact, the region’s supposed inhospitableness, together with its economic
backwardness, led Celso Furtado (1959; 1969), the architect of the region’s first comprehensive
development strategy, to propose “the reduction of the overpopulation in the semi-arid (…), by means
of migrating people to (…) the coastal regions”47. However, the people strongly opposed (Gomes,
2001), and the backlands experienced an average annual population growth of 1.8% from 1970 –
2000.48,49
45
Souce: Instituto Brasileiro de Geografia e Estatística (IBGE).
This particular data is from the 1987/1988 drought (Gomes, 2011).
47
Translated by the author from pp. 32-33 in Furtado (1969)
48
Source: Instituto Brasileiro de Geografia e Estatística (IBGE).
49
Today, Brazil’s backlands are the only densely populated “semi-arid” region in the world (Unger, 2009).
46
117
Today, while the Northeast is experiencing rapid growth, mainly due to the vertiginous growth of the
coastal capitals (Visser, 2015), which have been injected with multi-billion public and private
projects, the backlands are stuck in a void between two “illusions of development” (Unger, 2013):
On the one hand, there is poverty alleviation, most saliently in the form of conditional cash transfer
programs like Programa Bolsa Familia. While having lifted 30 million Brazilians out of extreme
poverty (Soares et al, 2010), these programs, by themselves, do not pose a transformative, real
development model (Araújo and Lima, 2009). Or, in other words, they are a necessary condition for
development, but not a sufficient one.
On the other hand, the main industrial policy has been what Unger (2009) has called “São Paulism”:
the attempt of replicating the development path the richer Southeast went through in the 1960s 50,
through the subsidisation of large-scale public works and companies.51 However, with modernization
theory largely discredited, this approach no longer corresponds to a contemporary understanding of
economic development (Diamond, 1992; Kay, 2010).
The insufficiency of the current development programs heightens the importance of an alternative,
and, despite some isolated work52, a real alternative is yet to emerge.53
3. Towards a model of post-Fordism
Could a transformative development model be organized around post-Fordism? Before arguing for
precisely this, let me briefly explain the theory. Post-Fordism is a descriptive economic theory of the
forms of production that first emerged in the late 1970s54, and can be found today in the world’s
productive vanguards, such as Silicon Valley, Baden-Württemberg, or the “Third Italy” (Bianchi and
Gualtieri, 1990; Rajan and Zingales, 2001). It differs substantially from its historic predecessor,
Fordism55, which refers to the forms of production found in mass production industries, which in turn
replaced traditional craft production (pre-Fordism) as the hegemonic form of production in the
1920s.56
By recombining the central literature, I derive the following model of post-Fordist forms of
production, which I summarize in Fig. 3.
50
This was part of a wider economic development model, Import Substitution Industrialization, which was the hegemonic approach to
development in Latin America in the 1960s (Baer, 1972).
Which was central to the then hegemonic model of Import Substitution Industrialization (Prebisch, 1967).
51
For example, the R$5bn investment by Italian car manufacturer Fiat is presented by the state’s development agency, the AD Diper
(Agência de Desenvolvimento do Estado de Pernambuco), as the largest success case (AD Diper, 2015).
52
There has been some work on the economic value of clusters all around the Northeast, trying to identify the reasons for their success. See,
for example, Apolinário and Silva (2009) and Costa (2011).
53
Unger (2009) has been considered to be the second comprehensive development model for the Northeast, after that of Celso Furtado
(Teixeira, 2015). My undergraduate thesis is a contribution towards the larger model proposed by Unger. However, despite the publishing in
Unger (2009), real action is yet to be taken, and the model is yet to be substantiated through public policy.
54
First emerged in productive vanguards in Japan, Europe, and the United States (Piore and Sabel, 1984), as researchers noticed that there
were high-growth economic clusters that did not have large-scale, mass production corporations
55
Fordism takes its name from Ford Motor Company and Henry Ford’s famous production line, which is considered by many to be the first
instance of industrial mass production and standardization of goods (Amin, 2011).
56
See, for example, Piore and Sabel (1984) or Best (1990).
118
Figure 3: Basic notions when comparing models of Fordism and post-Fordism
Firstly, on the worker level, one does not observe a rigid division of labour in post-Fordism (Best,
1990), as there is a blurring of lines between supervisory and executive roles, between employers and
shop-floor workers (Lazonick, 1990). Workers are skilled, and work on flexible equipment, as their
tasks involve creative processes that cannot be performed by a machine (Piore and Sabel, 1984).57
This is often associated with an entrepreneurial spirit within workers (Boltanski and Chiapello, 1999),
who, further, seek to use their potential to influence the quality and nature of produced goods, as
opposed to the quantity only (Best, 1990; Sturgeon, 2003).
Secondly, on the firm level, “flexible specialization” (Piore and Sabel, 1984) is central to postFordism, as firms equip themselves with multi-use equipment, seeking continuous improvement in
both products and process (Best, 1990). If market conditions change, either due to changing
preferences or technological advances, companies adapt, and are not locked into obsolete technologies
and processes (Saxenian, 1994). Production is focussed on permanent innovation of individualised
goods (Unger, 2007), as opposed to speed optimization in the production of large quantities of
standardized goods (Tolliday and Zeitlin, 1986). Furthermore, firms compete in the realm of
innovation, as opposed to cost minimisation (Piore and Sabel, 1984).
Thirdly, the relationship between companies is governed by “cooperative competition” (Piore and
Sabel, 1984) as the lines between cooperation and competition are blurred (Unger, 2007). Mostly
small, independent firms, while competing in the same or a similar sector, pool resources and provide
mutual services to one another (Best, 1990; Sturgeon 2003). They do so in order to achieve the
economies of scale required to compete with the vertically-integrated Chandlerian corporation 58
which are characteristic of Fordism (Lazonick, 2010). Moreover, the relationship between
government and firms is one of “strategic collaboration” (Rodrik, 2008), with the aim of jointly
identifying roadblocks and arriving at solutions.
More recently, the literature has hinted towards the existence of pre-Fordist (as opposed to Fordist)
conditions facilitating the emergence of post-Fordism (Bresnahan et al 2001; Berger, 2012). Silicon
Valley, for example, never ran through Fordism, and its pre-Fordist conditions of thick associative
networks, a culture of creativity, collaborative ties between the private and public sectors, as well as
small-scale production of individualised goods, seemed to have encouraged the emergence of postFordism (Saxenian, 1994; Sturgeon, 2003). While not conclusive, it appears that a region “does not
57
58
Unger (2014) describes this as production becoming “the embodiment of experimental thinking.”
As first proposed by Alfred Chandler (1993).
119
have to go through the purgatory of belated Fordism, but can rather move directly from pre- to postFordism”.59 I observe exactly this phenomenon in the region of study (see below).
4. Methodology
After introducing the socio-economic situation in the backlands, as well as presenting a model of
post-Fordist production, the task is now to combine both. The objective is to assess whether a
development model geared around post-Fordism could be appropriate. The very limited availability of
reliable data of good quality, together with my research question inherently being a qualitative one,
required a methodology that is unconventional in contemporary economics: primary research. I chose
a textile cluster, formed out of 3 towns and 7 villages in the backlands of the state of Pernambuco
(Fig. 4), as the area to conduct fieldwork in (Fig. 5). This cluster is considered to be representative for
dozens of further, similar clusters in the backlands (Apolinário at al, 2011), while Unger (2009; 2014)
has also pointed towards the suspected existence of post-Fordism there. I spent 6 weeks in the area,
carrying out interviews with local government, academics, development and governmental agencies,
trade unions, as well as entrepreneurs, employers and shop-floor workers in over 50 SMEs. I
conducted in-depth conversations interviews with over 200 individuals in all 10 towns/villages.
Cumulatively, I spent over 170 hours visiting SMEs and observing patterns of production. While
unorthodox in its method, this has been the only way for me to gain an in-depth understanding of the
economic processes at hand, in order to compare how far the attributes of post-Fordism outlines in my
model are observed in reality. Wherever possible, I carefully complement my observations with
reliable data.60
Figure 4: Area of Research61
Figure 5: Conducting grassroots research
Notes: The stars depict the location of the 10 villages
Source: Author’s own elaboration; Google Maps
Source: Author’s own photograph
59
See, for example, Unger (2014). It seems that post-Fordism is much closer related to pre-Fordism than Fordism, and this, areas in which
mass production never occurred, can transition directly to post-Fordism. Piore and Sabel (1984) remark that in the “Third Italy”, where there
were strong networks of craft production, post-Fordism flourished, and not in the industrially most advanced zone of North Italy.
60
The main datasets I use are from Brazil’s census in the IBGE, and from the SEBRAE-PE (Serviço Brasileiro de Apoio aos Micro e
Pequenos Emprendimentos, Pernambuco) database, who conducted two large surveys and interviews in 2002 and 2012 in the region of
study.
61
The towns and villages at the centre of this research are: Agrestina, Brejo da Madre de Deus, Caruaru, Cupira, Riacho das Almas, Santa
Cruz do Capibaribe, Surubim, Taquaritinga do Norte, and Toritama.
120
5. Results and Discussion
The fact that my area of study surpassed Brazil, the Northeast, and Pernambuco in both real GDP
growth and in population growth (Fig. 6) is an indication that some economic phenomenon has been
present, which I will now argue to be a transition to post-Fordism.
60%
50%
40%
30%
20%
10%
0%
Real GDP Growth 2002 - 2009
Area of
Study
56.1%
Population Growth 2002 - 2009
27.1%
36.2%
Northea
st
47.9%
Pernam
buco
44.3%
12.3%
11.2%
11.1%
Brazil
Figure 6: Comparative Real GDP Growth and Population Growth, 2002 – 2009
Source: IBGE: Censos Demográficos 2000/2010. Ipedata: Calculation of GDP at constant prices.
I divide the results from my fieldwork into three categories: firstly, attributes that correspond to postFordism; secondly, attributes that are in transition from pre- to post-Fordism; thirdly, crucial elements
for post-Fordism that are still absent.62 I summarise these in Fig. 7.
Figure 7: Matrix of attributes and stages of post-Fordism
Source: Author’s own elaboration
Post-Fordist attributes
Surprisingly, I observe a series of post-Fordist attributes in the region already. These are found mostly
on the individual worker level. Firstly, the individuals I interviewed, women and men alike, are selfmade individuals (Bordieu, 1998): without significant help from third parties, and in the absence of
adequate socio-economic conditions, have they managed to prosper in relative terms, “lifting”
themselves out of poverty through hard work.63 82% of the interviewees worked overtime at least
62
I have made the conscious decision to present my results in these categories, in order to most directly answer my research question.
Within these categories, as the astute reader will notice, one can find the categories that I divided the model of post-Fordism into, too. To
illustrate this, I have used these two categories as the axes of the matrix in Figure 7.
63
Fig. 8 shows a photograph of myself with one such individual that I met during my fieldwork.
121
twice per week, 73% had a part-time employment or their own businesses next to their full-time job,
and 55% self-identified as engaging in night-time study of either vocational or academic nature.64,65
Secondly, the entrepreneurial spirit in the region is clearly visible: there are over 60,000 active
entrepreneurs (given a population of 650,000), and many of those who are “only” shop-floor workers
have concrete plans to engage in entrepreneurial endeavours, mostly of creative nature. This element
is the third key observation: the main driver of the people’s entrepreneurial spirit appears to be an
expression of creativity and innovation as means of self-fulfilment and vocation66, magnified due to
the creative nature of the textile industry.
Furthermore, I also identify important elements of post-Fordism on the inter-firm level. The central
one is collaborative competition, a widespread practice among firms in the region. Given the size of
the companies (see below), this is crucial for them to compete with international and national
competition. In Toritama, despite 85% of companies specialising in denim-wear, they constantly pool
resources and provide services to one another. They even created the NCTPE67, an open organization
with the aim of fostering collaboration among competing companies with porous boundaries.
Fig. 8: Self-made entrepreneur in Caruaru
Source: Author’s own photograph
Pre- to post-Fordist transition attributes
It is mostly on the intra-firm level that I observe a direct transition: while some firms engage in
vanguard production processes68, and could be situated in any of the world’s productive vanguards,
the majority of firms mostly demonstrate pre-Fordist attributes.
Firstly, in terms of size: from the 18,803 firms registered in the SEBRAE database, 73% have up to 2
employees, 12% have 5 or more employees, and 3% have more than 15 employees. The majority of
“micro” companies are very rudimentary, often even operating out of the owner’s house, whereas
there exists an increasing number of well-established “medium-sized” firms.69
64
Numbers based on the 212 interview observations conducted by myself. I would like to refer the reader to my undergraduate thesis for
more information and a discussion about selection bias.
65
My observations confirm Souza’s (2010) characterization of what he calls Brazil’s “battlers”: a new middle class that has emerged from
poverty through dedicated work.
66
This is what Boltanski and Chiapello (1999) call the “spirit” of post-Fordism.
67
Núcleo de Cadeia Têxtil e de Confecções do Estado de Pernambuco (Nucleus of the Textile Sector in the State of Pernambuco).
68
This included processes of business administration, and not just the physical production of products.
69
In 2002, it was only 203 businesses that had >15 employees (SEBRAE-PE Database).
122
Secondly, the “blurring of roles” is clearly observable. Unlike in other textile industries, e.g. China’s
Henan province (Lin et al, 2011), where workers perform Fordist tasks of constant repetition, workers
in the research area partake in different tasks in both production and administration. In some firms
this is due to true vanguard practices, while in many it is simply due to the family running a micro
business. The transition is, however, clear: in 2002, family members worked in 83% of companies; in
2012, it was only in 68%.70
Thirdly, the most important company-specific attribute of post-Fordism is “flexible specialization”.
The textile industry’s fast-moving nature, due to constantly-changing tastes and trends, requires
vanguard firms to innovate permanently. To do so, they have equipped themselves with multi-use
machinery. From the 55,000 electronic machines in use in the area, about 25% (or about 14,000) are
considered to be of this type. The firms employing these are mostly those who have at least one
registered brand (22% of companies71). These firms can be considered to be of post-Fordist nature, as
they constantly seek to innovate and operate at the global frontier. At the same time, the majority of
firms using non-electronic, craft machines for the production of clothes without a own brand, do not
adapt to the changing market conditions swiftly.
Fig. 9: Craft production in Toritama
Fig. 10: Flexible specialization in Toritama
Source: Author’s own photographs
Source: Author’s own photographs
The transition from a craft based industry into a vanguard one can also be seen by the
formality/informality statistics 72 : in 2002, 92% of firms in Toritama and 94% in Santa Cruz do
Capibaribe were informal, while, in 2012, 77% were in Toritama and 81% in Santa Cruz do
Capibaribe.73
70
Simultaneously, the mean number of workers per firm from the direct family went down from 2.5 to 2.
The SEBRAE estimates that 36% of companies have a known brand, of which about 60% have at least one registered brand (5% have
more than one).
72
All vanguard producers I observed were formal, while all informal companies I observed pertained to the rearguard. This is, clearly, not
saying that there are no rearguard producers in the formal sector, and just an expansion of formality does not causally imply an increase in
formality. However, there appears to be a clear association between formality and vanguard forms of production, which is why I use this as
the best proxy.
73
The SEBRAE-PE database only has data for Toritama and Sta. Cruz for both 2002 and 2012; this data, however, is trustworthy.
71
123
Number of firms with >15 employees
Firms in which a direct family
member is an employee, in %
700
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
600
500
400
300
200
100
0
2002
2012
2002
Figure 11: “Medium-sized” firms in 2002/12
2012
Figure 12: Percentage of firms employing
family members
Source: SEBRAE-PE database
Source: SEBRAE-PE database
Informal firms, as % of total
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
2002
2012
Toritama
92%
77%
Sta Cruz
94%
81%
Figure 13: Informality in 2002/2012
Source: SEBRAE-PE database
Absent attributes
While there is a number of absent elements that most productive vanguard display, such as the
proximity of higher education centres or very good infrastructure, two missing elements are especially
worrying:
Firstly, the lack of basic education among a sizeable proportion of the population, as
about 20% of the working population is estimated to be analphabetic, and the SEBRAE estimates that
not even 30% of population aged 11 – 14 will successfully finish middle school. While vocational and
higher education options are emerging 74 , the access to basic education is fundamental for postFordism to flourish (Unger, 2013).
Secondly, there is very little public-private cooperation, as firms and the government hardly interact at
all. However, firms need the government, and the government needs the firms to identify problems
(Rodrik, 2008). Not only is the institutional framework unavailable, but the predisposition does not
seem very high either: 85% of respondent in the interviews I conducted did not want to be involved in
The state university, the UFPE (Universidade Federal de Pernambuco) opened a “backlands campus” in the 2000s, expanding the reach of
the public higher education system into the backlands.
74
124
any government-sponsored activities, and 63% rated the government’s involvement as “bad” or “very
bad” in the SEBRAE-PE survey.
When putting the individual pieces back together, what do I conclude? The results indicate that the
region is experiencing a clear transition to post-Fordism, with many crucial post-Fordist attributes
already present, while many elements are also currently in the transition period. At the same time,
pivotal elements (education, public-private cooperation) are missing outright, posing a substantial
threat for transformative, sustainable development. It is also important to note that my area of
research, despite there being dozens of comparable areas scattered around the Northeast, is a relative
success story, and many areas are much further behind; hence, relying even more heavily on the
provision of public policy to support the transition.
6. Policy Implications
Having discussed the main results, the question is now what to do about them. Firstly, I will outline
some high-level policy guidelines that need to be followed. I argue that there is a set of twin
considerations that need to be fulfilled. This consists of a reinterpretation of industrial policy for postFordism, as well as the provision of functional equivalents to the currently absent conditions. Both are
vital for bringing the vanguard to the rearguard.
The people and arising entrepreneurs in the backlands need to be given the means and opportunities to
grow and empower themselves economically. This requires the government to leave the tradition
behind of picking certain sectors to create national champions through subsidization and protection
(Hausmann and Rodrik, 2006). Instead, the focus should be the expansion of access to credit,
technology, and R&D for SMEs (Unger, 2007), with the state supporting a climate in which
innovation-friendly practices can flourish (Hausmann et al, 2008).
Furthermore, the new post-Fordist reality requires innovation in the institutional arrangements
between firms, as well as between firms and government (Weiss, 2013). On the horizontal axis,
competitive cooperation between firms needs to be encouraged, supported with an institutional
framework (Unger, 2007). On the vertical axis, participatory and decentralised legal institution need
to encourage strategic collaboration between the government and the private sector (Hausmann et al,
2012), as this interactive process elicits information on business opportunities and constraints,
generating policy responses (Rodrik, 2006)75.
I will now move, in a second instance, to illustrate the general idea with two proposals of concrete
public actions. The list is open and exemplary, and the proposals should have immediate practical
value, while also paving the way for a long-lasting transformative development path.76
First proposal: Equip the SEBRAE with a technology and R&D branch
The SEBRAE is an island of bureaucratic excellence in the backlands, performing very high quality
work in supporting businesses and transferring knowledge in the administration branch. The
institution is very highly regarded among the vast majority of entrepreneurs, employers, and shopfloor workers I interviewed. Technological backwardness and lack of own R&D capabilities are
among the main problems that firms in the backlands face to innovate, and compete nationally.
Hence, it is pivotal to expand the access to technology and R&D for arising entrepreneurial firms. The
As put by Rodrik (2006): “industrial policy is a state of mind”, and is not refined to arm’s length regulation by the government or topdown state capitalism. And while this is certainly innovative it is not revolutionary, as the world’s productive vanguards have such a model
in operation (Bresnahan et al, 2001).
76
The list is furthermore not exhaustive, as fundamental areas, such as the renovation of agriculture and the irrigation question are not even
touched upon. For a broader set of proposals, the reader should refer to my undergraduate thesis.
75
125
SEBRAE should be further equipped with a technology and R&D branch, in order to expand access to
these to more firms, so that they can transfer from pre- to post-Fordism. This can either be done via an
extension of the SEBRAE, by means of equipping the current service with a broader mandate or
creating new SEBRAE divisions (Campos et al, 2011), or by way of a new institution established
through the cooperation of several relevant institutions, including the SEBRAE as main stakeholder
(Unger, 2009).77
Second proposal: Expand the access to credit by eliminating bureaucratic roadblocks and
fulfilling the BNDES78 mandate
87% of companies in the region of study have never requested a loan from a financial institutions, but
only 19% reported that this was due to a lack of a supply by retail banks operating in the area. 79 The
main problems are legal requirements that only few companies comply with, such as the bureaucratic
and relatively hard to acquire CNPJ. 80 These bureaucratic roadblocks result in firms resorting to
insecure and informal non-bank financing when in need of credit. Expanding access to formal credit is
central for the growth of firms, as well as reduction of informality. A two-fold approach is needed: on
the one hand, bureaucratic roadblocks should be eliminated to expand the access to credit (such as a
simplification of CNPJ process), while, on the other hand, the BNDES mandate, which includes the
provision of credit to SMEs, should be fulfilled, and thus launching the process of bureaucratic
reform.81
I encourage the interested reader to refer to my undergraduate thesis for a further set of policy
proposals, addressing the problems of collaborative competition, strategic cooperation, poor
education, and predominance of informality.82
7. Conclusion
Let me conclude this research essay by reiterating my central claim: a transformative development
model geared around post-Fordism is adequate and feasible for Brazil’s backlands. Through the
analysis of my fieldwork, I have shown that many post-Fordist elements are already present in the
backlands today, with others being in a direct transition from pre- to post-Fordism. Nevertheless,
some pivotal attributes are completely absent. This is where public action needs to come in, in order
to support the direct transition and establish functional equivalents to encourage post-Fordism. If
executed well, the region, considered to be Brazil’s rearguard, could develop into one of the country’s
vanguards: the backlands have the unique opportunity to take the lead in becoming a regional,
experimental alternative to the status quo.
Many people followed-up on the question, with which I began this paper’s introduction, with the
following question: “Este modelo de desenvolvimento transformador se tornará realidade?” (“Will
this transformative development model become reality?”). There are signs for contained optimism.
Roberto Mangabeira Unger was appointed as Brazil’s Minister of Strategic Affairs in February 2015
77
Such as the Federal Government (the Ministry of Science and Technology), the SENAI (the National Service for Industrial Learning), and
the FINEP (the Financier of Studies and Projects for the Northeast).
78
Banco Nacional de Desenvolvimento (Brazilian Development Bank)
79
Data from SEBRAE-PE database. Interest rates are further said to be competitive, with banks such as the Banco do Brasil, the Caixa
Econômica, and the Banco do Nordeste, operating in the region.
80
The CNPJ is a number given out by the registry of firms, and is time-consuming and inefficient to acquire. As a result, most firms do not
have it in the backlands, which immediately disqualifies them from loans, despite, financially, being eligible to get them.
81
The BNDES main occupation has been the financing of M&A deals for Brazilian champions, despite having a budget which is 3x larger
than that of the World Bank (Unger, 2014). Its mandate, however, clearly states that the provision of credit to SMEs in under-developed
region is one of its main tasks (Koteski, 2004; Morais, 2008). But it does not provide credit in the backlands, which falls under its mandate.
82
Including: Third Proposal: Reduce informality by educating micro and small firms, as well as simplifying the formalization process;
Fourth Proposal: Strengthen and fund sector associations, such as the NCTPE, to foster collaborative competition; Fifth Proposal: Create
“windows” of public-private strategic collaboration; Sixth Proposal: Increase the access to middle schooling for all children in the
backlands; Seventh proposal: Institute distance learning mechanisms.
126
by president Dilma Rousseff. This provides an enormous political impulse to the socio-economic
movement occurring in the region today and to the potential for transformative development. Even if
only incrementally, I hope that my research has contributed towards an alternative development
model for the backlands, by further uncovering the transformative socio-economic phenomenon of the
emergence of post-Fordism. Indispensably, substantial further research needs to be carried out, and
problems concerning the agricultural sector and the omnipresent “water problem” need to addressed
in order to provide the region with the prospect of a brighter future.
Bringing the vanguard to the rearguard is, without doubt, a very ambitious project. But if the past 12
months that I have spent working on this project have taught me something, then it is that this is not
utopia or a vision of a distant future, but that it can be done, and it can be done now.
References
Amin, Ash. 2011. Post-Fordism: A Reader. New York: John Wiley & Sons.
Apolinário, Valdênia and Silva, Maria Lussieu da. 2009. “Organização da produção e processo de
trabalho em áreas criativas: o APL de bordado de Caicó/RN.” Revista ABET, 8 (2), pp. 7-29.
Apolinário, Valdênia; Silva, Maria Lussieu da and Garcia, Odair Lopes. 2011. Análise do
Mapeamento e das Políticas para Arranjos Produticos Locais no Norte, Nordeste e Mato Grosso e
dos Impactos dos Grandes Projetos Federais no Nordeste. Rio de Janeiro: BNDES.
Araújo, Leonardo Alves de and Lima, João Policarpo R. 2009. “Transferências de renda e
empregos públicos na economia sem produção do semiárido nordestino.” Planejamento e políticas
públicas, 33 (1), pp. 45 – 77.
Baer, Werner. 1972. “Import substitution and industrialization in Latin America: experiences and
interpretations.” Latin American Research Review, 7 (1), pp. 95-122.
Berger, Suzanne. 2012. “Toward a third industrial divide?” in Paul Osterman, ed., Economy in
society: Essays in honor of Michael J. Piore. Cambridge, MA: MIT Press, pp. 54-79.
Best, William H. 1990. The new competition: institutions of industrial restructuring. Cambridge,
MA: Harvard University Press.
Bianchi, Paolo and Gualtieri, Guido. 1990. “Emilia-Romagna and its industrial districts: the
evolution of the model” in Leonardi, Robert and Nanetti, Raffaella, eds., The regions and European
integration: the case of Emilia Romagna. London/New York: Pinter Publishers, pp. 83 – 108.
Boltanski, Luc and Chiapello, Eve. 1999. Le nouvel esprit du capitalisme. Paris: Gallimard.
Bordieu, Pierre. 1998. Practical reason: On the theory of action. Stanford, CA: Stanford University
Press.
Bresnahan, Timothy; Gambardella, Alfonso and Saxenian, AnnaLee. 2001. “‘Old Economy’
Inputs for ‘New Economy’ Outcomes: Cluster Formation in the New Silicon Valleys.” Industrial and
Corporate Change, 10 (4), pp. 835-860.
Campos, Teodoro Malta; Lima, Edmilson de Oliveira; Urbanavicius Jr., Vladas and Alves,
Danilo Vieira. 2011. “Internacionalização de micro e pequenas empresas: um estudo de caso sobre a
actuação do Sebrae.” Revista da Micro e Pequena Empresa, 5 (1), pp. 36-50.
Chandler Jr., Alfred D. 1993. The visible hand. Cambridge, MA: Havard University Press.
Costa, Odorico D de Moraes Eloy da. 2011. Aranjos produtivos locais – APL’s como Estrtégia de
Desenvolvimento: Uma Abordagem Teórica. Fortaleza: IPECE.
127
Diamond, Larry. 1992. “Economic development and democracy reconsidered.” American
behavioral scientist, 35 (4-5), pp. 450 – 499.
Furtado, Celso. 1959. A operação Nordeste. Ministério da Educação e Cultura. Rio de Janeiro:
Instituto Superior da Educação e Cultura.
_______. 1967. Um política de desenvolvimento econômico para o Nordeste. Recife: SUDENE.
Gomes, Gustavo Maia. 2001. Velhas Secas em Novos Sertões. Brasília: Instituto de Pesquisa
Econômica Aplicada.
Hausmann, Ricardo and Rodrik, Dani. 2006. “Doomed to Choose: Industrial Policy as
Predicament.: Center for International Development at Harvard University Discussion Paper,
September, Harvard University.
Hausmann, Ricardo; Rodrik, Dani and Sabel, Charles. 2008. “Reconfiguring Industrial Policy: A
Framework with an Application to South Africa.” CID Working Paper No. 168, Harvard University.
Kay, Cristobal. 2010. Latin American Theories of development and underdevelopment. London:
Routledge.
Khan, Ahmad S. and Campos, Robério T. 1995. “Efeitos das Secas no Setor Agrícola do
Nordeste,” in Gomes, Gustavo Maia; Souza, Hermino Ramos de and Magalhães, Antonio Rocha, eds.,
Desenvolvimento sustentável no Nordeste. Brasília: Instituto de Pesquisa Econômica Aplicada, pp.
45-98.
Koteski, Marcos Antonio. 2004. "As micro e pequenas empresas no contexto econômico brasileiro."
Revista FAE Business, 8, pp. 16-18.
Lazonick, William. 1990. Competitive advantage on the shop floor. Cambridge, MA: Harvard
University Press.
_______. 2010. “The Chandlerian corporation and the theory of innovative enterprise.” Industrial and
Corporate Change, 19 (2), pp. 317-349.
Leskovec, Jure; Lang, Kevin J.; Dasgupta, Anirban and Mahoney, Michael W. 2009.
“Community structure in large networks: Nature cluster sizes and the absence of large well-defined
clusters. Internet Mathematics, 6 (1), pp. 29-123.
Lin, Hui-Lin; Li, Hsiao-Yun and Yang, Chih-Hai. 2011. "Agglomeration and productivity: Firmlevel evidence from China's textile industry." China Economic Review, 22 (3), pp. 313-329.
Morais, José Mauro de. 2008. "Programas especiais de crédito para micro, pequenas e médias
empresas: BNDES, Proger e Fundos Constitucionais de financiamento" in Negri, João Alberto de, ed.,
Políticas de Incentivo à Inovação Tecnológica. Brasilia: Instituto de Pesquisa Econômica Aplicada,
pp. 389-433.
Piore, Michael and Sabel, Charles. 1984. The second industrial divide: possibilities for prosperity.
New York: Basic Books.
Rajan, Raghuram and Zingales, Luigi. 2001. “The Firm as a Dedicated Hierarchy: A Theory of the
Origins and Growth of Firms.” The Quarterly Journal of Economics, 116 (3), pp. 805-851.
Rodrik, Dani. 2008. “Industrial Policy: Don’t ask why, ask how.” Middle East Development Journal,
1, pp. 1-29.
_______. 2009. One economics, many recipes: globalization, institutions, and economic growth.
Princeton: Princeton University Press.
128
Saxenian, AnnaLee. 1996. Regional Advantage: Culture and Competition in Silicon Valley and
Route 128. Cambridge, MA: Harvard University Press.
Soares, Fábio Veras; Ribas, Rafael Perez and Osório, Rafael Guerreiro. 2010. “Evaluating the
impact of Brazil’s Bolsa Familia: cash transfer programs in comparative perspective.” Latin American
Research Review, 45 (2), pp. 173-190.
Souza, Jessé. 2010. Os Batalhadores Brasileiros: Nova classe média ou nova classe trabalhadora?
Belo Horizonte: Editora da UFMG.
Teixeria, Carlos Sávio. 2015. “Os Projetos de Transformação do Nordeste: De Celso Furtado a
Roberto Mangabeira Unger.” Discussion Paper w/n, University of São Paulo.
Tolliday, Steven and Zeitlin, Jonathan. 1986. Between Fordism and flexibility. Oxford: Oxford
University Press.
Unger, Roberto Mangabeira. 2007. Free Trade Reimagined: The World Division of Labour and the
Method of Economics. Princeton: Princeton University Press.
_______. 2009. “O desenevolvimento do Nordeste como projeto nacional.” Brasília: Secretaria de
Assuntos Estratégicos.
_______. 2014. “Progressive Alternatives in the BRICs”. Lecture in Government 1092, Harvard
University, 13 March.
Visser, Ricardo. 2015. “As transformações do Nordeste em numerous e estatísticas.” Working paper
w/n, UFMG.
Weiss, John. 2013. “Industrial Policy in the Twenty-First Century: Challenges for the Future” in
Szirmai, Adam; Naudé, Wim and Alcorta, Ludovico, eds. Pathways to industrialization in the twentyfirst century: new challenges and emerging paradigms. Oxford: Oxford Universtity Press, pp. 373391.
129
Addressing the Exchange Rate Puzzle: Forecasting with a Taylor rule
inspired model by Amy Louise Curry
Final year, Economics with a Year Abroad
Fundamental exchange rate forecasting has proven extremely difficult since the break down of the
Bretton Woods system in the 1970’s. The disappointment climaxed with the publication by Meese
and Rogoff (1983), who showed that a simple random walk model dominated fundamental exchange
rate models in terms of predictive power; thus inferring that economic theory is irrelevant to exchange
rate modeling. This paper overthrows Meese and Rogoff’s claim by providing overwhelming
evidence that a Taylor rule fundamental based exchange rate model outperforms the random walk
model at the 1-month time horizon for all three currency pairs: USDJPY, USDDEU and USDGBP.
Further it provides evidence that an unexpected movement in the central bank interest rate acts a good
predictor of exchange rate movements for two out of three currency pairs: USDDEU and USDGBP.
Chapter One: A brief introduction into the literature surrounding exchange rate
forecasting
1.1 The Exchange Rate Puzzle:
The highly cited paper “Empirical Exchange Rate Models of the Seventies” by Meese and Rogoff
(1983) marked a cornerstone in exchange rate modeling. Meese and Rogoff compared the accuracy of
out-of-sample predictions of time series and structural models for major dollar spot rates. The key
finding states that naïve random walk model performs at least as well as any other model that runs on
economic fundamentals in terms of RMSE.
Following publication, many academics have gone on to reference Meese and Rogoff as the
benchmark in exchange rate forecasting; a testament to the paper being hailed as the “gold standard”
in such discussions. Such a bold claim has grown credibility as others have challenged it and failed;
overthrowing conventional macroeconomic models.
1.2 The Meese and Rogoff Puzzle: Reality or Myth?
It is no surprise that a wealth of academic journals have taken the stance of devil’s advocate;
questioning this bold consensus that a random walk model is one of superiority in exchange rate
forecasting.
A number of studies have looked at the exchange rate predictability using the uncovered interest
parity (UIP) condition. Campbell and Clarida (1987) investigate the relationship between the
exchange rate vis-à-vis the dollar and real interest rates. Campbell and Clarida’s findings suggest that
interest rate differentials are not sufficient to account for changes in the exchange rate. Likewise
Edison and Pauls (1993) state that although the UIP condition is theoretically logical, they have failed
to find evidence to suggest that the relationship between the interest rate and exchange rate
movement’s hold.
On the basis of the evidence currently available the UIP condition does not appear to hold, at least in
the short run. In modern exchange rate economics a multitude of papers have instead attempted to
build a two-country model that: (i) models the interest rate using the Taylor Rule and (ii) then
combine this with the UIP to build an overall model that predicts exchange rates. Molodtsova and
131
Papell (2009) find short-term out-of-sample predictability of models that incorporates Taylor Rule
fundamentals. In fact they find strong evidence of exchange rate predictability at the one month of 11
out of 12 OECD currencies vis-a-vis the US dollar. Likewise Engel and West (2006) construct a twocountry model that finds a positive correlation between the model rate and the realized dollar-mark
exchange rate. Furthermore Galimberti and Moura (2013) use a parallel approach applied to fifteen
emerging economies and they find strong evidence of exchange rate predictability using Taylor Rule
fundamentals.
In addition the role of incoming news is found to be significant in exchange rate forecasting. Evans
and Lyon (2007) suggest that macroeconomic news can account for approximately 30% of daily
exchange rate movements. Such a conclusion was made through utilizing a model that incorporates a
broad range of macroeconomic news; including news that is used on FX trading desks; to observe
how this influences the exchange rate. Andersen, Bollerslev, Diebold and Vega (2002) use data on
macroeconomic expectations and macroeconomic news to calculate surprises. They find that surprises
translate into significant differences in expectations to what is then realized consequently causing the
exchange rate to move.
Chapter Two: Constructing the exchange rate model
In this paper I intend to build a model that combines the exchange rate theory across the literature that
has been covered briefly in the literature review. The model will combine both a fundamental and
technical approach to exchange rate forecasting.
It is clear that the traditional UIP condition alone is not a strong enough predictor of exchange rates;
modern papers have had greater success in using models that combine the UIP condition and the
Taylor Rule (Molodtsova and Papell (2009), Engel and West (2006), Galimberti and Moura (2013)). I
will, therefore, for the fundamental part of my model use the UIP condition that incorporates Taylor
Rule fundamentals. In addition the model will incorporate a surprise announcement element. For the
technical component of my forecasting model will be the simple random walk model.
The exchange rate model proposed in this paper consists of four stages. The first stage introduces the
determination of the central bank interest rate as an endogenous monetary policy Taylor rule.
Secondly the model uses the uncovered interest parity relationship to connect movements in exchange
rate movements to interest rate differentials; thus linking the first stage to exchange rate movements.
As a third step a surprise macroeconomic news element will be incorporated. Lastly the model will
introduce a technical model widely known as a random walk; an AR(1) process with a coefficient
equal to one. The first two stages of the model used in this paper use Molodtsova’s and Pappell’s
(2009) finite-difference Taylor rule forecasting model.
My two hypotheses are as follows:
1. The Taylor rule fundamental based forecasting model is superior to that of the random walk
model
2. The Taylor rule fundamental based forecasting model that incorporates the surprise element is
superior to that of the basic Taylor rule fundamental based model
2.1 Stage One: Modeling the Taylor Rule
Since the 1980’s central banks have used interest rates as a policy instrument rather than controlling
aggregate measures of money supply. Taylor (1993) proposed the following model:
𝑖𝑑∗ = π‘Ÿ ∗ + πœ‹π‘‘ + πœ™(πœ‹π‘‘ − πœ‹ ∗ ) + 𝛾𝑦𝑑 (1)
132
where 𝑖𝑑∗ is the target for the short-term nominal interest rate, π‘Ÿ ∗ is the equilibrium real interest
rate, πœ‹ ∗ is the central banks inflation objective, πœ‹π‘‘ is the current period inflation rate and 𝑦𝑑 is the
current output gap.
Numerous papers that employ Taylor rule fundamentals allow for the target interest rate to be reached
gradually over time; this is otherwise called interest rate smoothing. This can be implemented by
introducing a weighted sum of the short-term nominal interest rate target and a lagged variable of the
short-term nominal interest rate:
𝑖𝑑 = (1 − 𝜌)𝑖𝑑∗ + πœŒπ‘–π‘‘−1 + πœπ‘‘ (3)
where 𝜌 is the weight that represents the central banks preferences to interest rate smoothing and πœπ‘‘ is
an error term.
By substituting (2) into (3) we obtain the following specification:
𝑖𝑑 = (1 − 𝜌)(π‘Ÿ ∗ + πœ‹π‘‘ + πœ™(πœ‹π‘‘ − πœ‹ ∗ ) + 𝛾𝑦𝑑 + π›Ώπ‘žπ‘‘ ) + πœŒπ‘–π‘‘−1 + πœπ‘‘ (4)
where 𝛿 = 0 for the United States.
To simplify this model the parameters πœ‹ ∗ and π‘Ÿ ∗ can be combined into a constant term πœ‡. This leads
to the following equation:
𝑖𝑑 = (1 − 𝜌)(πœ‡ + πœ†πœ‹π‘‘ + 𝛾𝑦𝑑 + π›Ώπ‘žπ‘‘ ) + πœŒπ‘–π‘‘−1 + πœπ‘‘ (5)
where πœ‡ = πœ‹ ∗ + π‘Ÿ ∗ and πœ† = 1 + πœ™.
Equation (5) describes how the central bank sets the nominal short-term interest rate today as a
function of inflation rates, output, real exchange rates and a lagged value of the real exchange rate.
2.2 Stage Two: Interest Rate Fundamentals
By the late nineteenth century it had become common knowledge among policymakers that the
behavior of exchange rates could be influenced through the adjustment of interest rates. By raising
domestic interest rates, the foreign exchange value of the domestic currency unit could be
strengthened. This provides the basis of the UIP condition that states the difference in interest rates
between two countries is equal to the expected change in exchange rates between the countries'
currencies.
However, on the basis of the evidence currently available the UIP condition does not appear to hold,
at least in the short run. In fact evidence suggests the entire opposite of the UIP condition, also known
as the forward premium puzzle. Therefore the model in this paper will include a more flexible
specification that has been used by Molodstova and Papell (2009):
Δ𝑠𝑑+1 = 𝛼 + πœ”(𝑖𝑑 − 𝑖̃)
𝑑 (7)
The variable st is the U.S. dollar nominal exchange rate determined as foreign price of a domestic
currency so that an increase in st is an appreciation of the dollar. The specification does not restrict
πœ” = −1; when πœ” = −1 this is consistent with the UIP condition. Furthermore the model does not
restrict the sign; when πœ” = 1 the specification is consistent with the forward premium puzzle.
133
To derive the forecasting equation, the interest rate differential must be constructed. From above the
Taylor rule interest rate specification is:
𝑖𝑑 = (1 − 𝜌)(πœ‡ + πœ†πœ‹π‘‘ + 𝛾𝑦𝑑 + π›Ώπ‘žπ‘‘ ) + πœŒπ‘–π‘‘−1 + πœπ‘‘ (5)
Therefore by subtracting the foreign countries interest rate equation from that of the US.
It follows that:
𝑖𝑑 − 𝑖̃𝑑 = 𝛼 + π›Όπ‘’πœ‹ πœ‹π‘‘ − π›Όπ‘“πœ‹ πœ‹
̃𝑑 +𝛼
𝑒𝑦 𝑦𝑑
+ 𝛼𝑒𝑦 𝑦𝑑 − 𝛼𝑓𝑦 𝑦̃𝑑 − π›Ώπ‘žΜƒπ‘‘ + πœŒπ‘’ 𝑖𝑑−1 − πœŒπ‘“ 𝑖𝑑−1
Μƒ + πœ‚π‘‘ (8)
where π›Όπœ‹ = πœ†(1 − 𝜌) and 𝛼𝑦 = 𝛾(1 − 𝜌) for both countries, and π›Όπ‘ž = 𝛿(1 − 𝜌) for the foreign
country.
By substituting (8) into (7) this constructs an exchange rate forecasting equation:
Δ𝑠𝑑+1 = 𝛼 + πœ” (𝛼 + π›Όπ‘’πœ‹ πœ‹π‘‘ − π›Όπ‘“πœ‹ πœ‹
̃𝑑 +𝛼
𝑒𝑦 𝑦𝑑
+ 𝛼𝑒𝑦 𝑦𝑑 − 𝛼𝑓𝑦 𝑦̃𝑑 − π›Ώπ‘žΜƒπ‘‘ + πœŒπ‘’ 𝑖𝑑−1 − πœŒπ‘“ 𝑖𝑑−1
Μƒ + πœ‚π‘‘ ) (9)
This can be simplified to the resulting empirical model:
Δ𝑠𝑑+1 = 𝛽0 + 𝛽1 πœ‹π‘‘ + 𝛽2 πœ‹
̃𝑑 + 𝛽3 𝑦𝑑 + 𝛽4 𝑦̃𝑑 + 𝛽5 π‘žΜƒπ‘‘ + 𝛽6 𝑖𝑑−1 + 𝛽7 𝑖𝑑−1
Μƒ + πœ‚π‘‘ (10)
2.3. Stage Three: Central Bank Surprises
Dornbusch (1976) analysed the process of adjustment to an unanticipated change in money supply,
demonstrating that the initial jump in the exchange rate exceeded the adjustment in the long-run
equilibrium. On the basis of the model proposed by Dornbusch and the subsequent literature on
overshooting, it seems fair to suggest that the exchange rate would immediately jump ex post a
surprise interest rate announcement.
The key aspect here is to introduce a central bank surprise element into the Taylor rule based
forecasting model. Introducing a variable for a surprise announcement presents a challenge in itself;
there is little if any data on surprise announcements for agents. For that reason I will improvise by
identifying unanticipated changes in interest rates. I will introduce a “surprise announcement”
indicator variable that will take the following values:
1: 𝑖𝑓 π‘π‘œπ‘ π‘–π‘‘π‘–π‘£π‘’ π‘ β„Žπ‘œπ‘π‘˜
0: 𝑖𝑓 π‘›π‘œ π‘ β„Žπ‘œπ‘π‘˜
𝑁𝑑 = {
−1: 𝑖𝑓 π‘›π‘’π‘”π‘Žπ‘‘π‘–π‘£π‘’ π‘ β„Žπ‘œπ‘π‘˜
The indicator will take the value one if there is a positive shock: an interest rate differential increase
by at least 150%. It will take the value of zero if there is no shock to the interest rate differential.
Finally the indicator will take the value of negative one if there is a negative shock: an interest rate
differential decline by at least 150%. In summary if there is a positive interest rate shock, where
π‘–π‘‘π‘ˆπ‘† > 𝑖𝑑𝐹 , then the dollar (foreign currency) will be expected to appreciate (depreciate). If there is a
negative interest rate shock, where 𝑖𝑑𝐹 > π‘–π‘‘π‘ˆπ‘† then the dollar (foreign currency) will be expected to
depreciate (appreciate).
To derive the full forecasting equation the newly introduced “surprise announcement” indicator
variable is included:
Δ𝑠𝑑+1 = 𝛽0 + 𝛽1 πœ‹π‘‘ − 𝛽2 πœ‹
̃𝑑 + 𝛽3 𝑦𝑑 − 𝛽4 𝑦̃𝑑 − 𝛽5 π‘žΜƒπ‘‘ + 𝛽6 𝑖𝑑−1 − 𝛽7 𝑖𝑑−1
Μƒ + 𝛽8 𝑁𝑑 + πœ‚π‘‘ (11)
134
2.4. Stage Four: The Technical Model
The technical approach to forecasting focuses on extrapolating past exchange rate trends. There is
overwhelming evidence suggesting that a random walk model distinctly dominates theoretical and
technical models in terms of predictive performance for the major dollar spot rates. It follows
conventionally that the majority of papers that incorporate Taylor rule fundamentals include a random
walk element into the corresponding forecasting model,
The forecasting model can now be updated to the following:
𝑠𝑑+1 = Δ𝑠𝑑+1 + 𝑠𝑑
𝑠𝑑+1 = 𝛽0 + 𝛽1 πœ‹π‘‘ + 𝛽2 πœ‹
̃𝑑 + 𝛽3 𝑦𝑑 + 𝛽4 𝑦̃𝑑 + 𝛽6 𝑖𝑑−1 + 𝛽7 𝑖𝑑−1
Μƒ + 𝛽8 𝑁𝑑 + πœ‚π‘‘ + 𝑠𝑑 (12)
Chapter Three: Estimation and Forecasting
3.1. Data
The primary source of data on macroeconomic fundamentals was collected from the OECD iLibrary
Database. I use monthly data on the short-term 3-month interest rates; the rates at which short term
borrowings are made between financial institutions. The price level in an economy is measured by the
monthly Consumer Price Index (CPI). The CPI is measured in terms of the annual growth rate and in
index, 2010 is the base year. As a proxy for national income I use the monthly industrial production
index, where 2010 is the base year. The reason for not using data on GDP is that it is only released on
a quarterly basis.
The currencies considered in this paper are the British pound, Japanese yen and the Deutsche mark.
The currency data is obtained from the Bank of England’s database. The models for the USDGBP
currency pair are estimated using monthly data from February 1978 to January 2015. The second
currency pair, the USDDEU, is estimated using data from February 1978 to December 2001. The
USDJPY currency pair is estimated using monthly data from May 2002 to January 2015. The reason
for the shorter data span for the USDDEU currency pair is due to Germany joining the European
Monetary Union in 2002. The USDJPY currency pair is also of a shorter span due to data
unavailability on the 3-month short-term interest rate before May 2002.The exchange rates are
defined as the foreign price of the U.S. dollar nominal exchange rate so that an increase in the
exchange rate is an appreciation of the dollar.
3.2. Forecasting
This paper constructs 1-month ahead forecasts for each of the three currency pairs: the USDJPY,
USDGBP and USDDEU. I use data for approximately half of data span in the estimation of the
respective linear regressions. The remaining data is then reserved for out-of-sample forecasting. I
forecast using both a pseudo (recursive) out-of-sample forecast and a rolling-window out-of-sample
forecast. A pseudo-out-of-sample forecast re-estimates the model every period in the out-of-sample
time period, allowing the model to be updated each time period to take account for data that has
occurred in recent periods. A rolling window forecast is one that has a fixed time window, in this case
60 observations, and then re-estimates the model each period for the out-of-sample period as it moves
the window through the sample.
I will estimate three Taylor rule forecasting models for each of the currencies, one with the indicator
of a surprise interest rate shock (at 150%), secondly one with the indicator of a surprise interest rate
135
shock (at 100%) and lastly a model that excludes the indicator for a surprise interest rate shock. These
are defined below:
Δ𝑠𝑑+1 = 𝛽0 + 𝛽1 πœ‹π‘‘ + 𝛽2 πœ‹
̃𝑑 + 𝛽3 𝑦𝑑 + 𝛽4 𝑦̃𝑑 + 𝛽6 𝑖𝑑−1 + 𝛽7 𝑖𝑑−1
Μƒ + 𝛽8 𝑁150%,𝑑 + πœ‚π‘‘
Δ𝑠𝑑+1 = 𝛽0 + 𝛽1 πœ‹π‘‘ + 𝛽2 πœ‹
̃𝑑 + 𝛽3 𝑦𝑑 + 𝛽4 𝑦̃𝑑 + 𝛽6 𝑖𝑑−1 + 𝛽7 𝑖𝑑−1
Μƒ + +𝛽8 𝑁100%,𝑑 + πœ‚π‘‘
Δ𝑠𝑑+1 = 𝛽0 + 𝛽1 πœ‹π‘‘ + 𝛽2 πœ‹
̃𝑑 + 𝛽3 𝑦𝑑 + 𝛽4 𝑦̃𝑑 + 𝛽6 𝑖𝑑−1 + 𝛽7 𝑖𝑑−1
Μƒ + πœ‚π‘‘
The identity to find the full predicted exchange rate is as follows:
𝑠𝑑+1 = Δ𝑠𝑑+1 + 𝑠𝑑
3.3. Evaluation of out-of-sample forecasts
RMSE:
The RMSE of a model prediction is defined as the square root of the mean squared error:
RMSE =
å
n
i=1
(Dsobs,t - Dsmo del,t )2
n
Directional Accuracy:
What is more important, particularly for businesses and the government, is the directional accuracy of
the exchange rate forecast. For this reason I created a dummy variable that reads one if the forecast
predicts the same direction of the exchange rate movement as the realized exchange rate movement
and zero if not. A percentage of correct direction predictions are then constructed across the models
and currencies.
Clark and West test statistic:
The Clark and West (CW) test statistic has become one the most popular out-of-sample test statistic in
exchange rate literature to test two nested models against each other. Clark and West (2007) develop a
test statistic that can be applied to standard normal critical values.
Model one is the null model and model two is the alternative model. First the following is estimated:
Μ‚ 2
Μ‚ 2
Μ‚
Μ‚ 2
𝑓̂̂
𝑑+1 = (𝑠𝑑+1 − 𝑠̂1𝑑,𝑑+1 ) − [(𝑠𝑑+1 − 𝑠̂2𝑑,𝑑+1 ) − (𝑠̂1,𝑑+1 − 𝑠̂2𝑑,𝑑+1 ) ]
Where 𝑠𝑑+1 is the realized exchange rate value, 𝑠̂̂
1𝑑,𝑑+1 is the predicted exchange rate value from the
model 1 and 𝑠̂̂
2𝑑,𝑑+1 is the predicted exchange rate value from the second model. The test statistic can
be calculated after 𝑓̂̂
𝑑+1 has been calculated as following:
π‘‡πΆπ‘Š =
√𝑃. 𝑓 Μ…
Μ…
π‘‰π‘Žπ‘Ÿ(𝑓̂̂
𝑑+1 − 𝑓 )
Where 𝑃 is the number of out-of-sample predictions and 𝑓 Μ… = 𝑃−1 ∑𝑇𝑑=𝑇−𝑃+1 𝑓̂̂
𝑑+1 .
Chapter Four: Empirical Results
3.3. Results
Unfortunately I was unable to estimate the Taylor rule based models that incorporated the surprise
element the USDJPY currency pair. This is because in the in-sample period there were no surprise
central bank announcements that caused a large enough differential to allow the indicator to equal
136
anything other than zero. Consequently for the USDJPY currency pair there is only the results for the
model that does not include the surprise variable. For the USDGBP and USDEU currency pairs the
surprise elements are incorporated.
3.3.1.
Directional Accuracy and RMSE
USDGBP
Random Walk
Surprise (150%) out-of-sample pseudo
Surprise (150%) out-of-sample rolling
Surprise (100%) out-of-sample pseudo
Surprise (100%) out-of-sample rolling
No Surprise out-of-sample pseudo
No Surprise rolling
RMSE
USDJPY
Random Walk
No Surprise out-of-sample pseudo
No Surprise rolling
RMSE
USDDEU
Random Walk
Surprise (150%) out-of-sample pseudo
Surprise (150%) out-of-sample rolling
Surprise (100%) out-of-sample pseudo
Surprise (100%) out-of-sample rolling
No Surprise out-of-sample pseudo
No Surprise rolling
RMSE
0.0206
0.0149
0.0160
0.0150
0.0162
0.0149
0.0160
Directional Accuracy
23.83%
51.06%
53.19%
51.49%
52.34%
51.91%
52.77%
3.5228
2.1444
3.0057
Directional Accuracy
58.44%
71.43%
57.14%
0.0689
0.0542
0.0616
0.0541
0.0611
0.0541
0.0612
Directional Accuracy
57.04%
53.52%
59.86%
54.23%
59.15%
53.52%
59.15%
For all three currency pairs it is clear that the Taylor rule based models overwhelmingly outperform
the random walk in terms Root Mean Squared Error (RMSE). Likewise in terms of directional
accuracy the random walk appears to be inferior. For the USDGBP currency pair the Taylor rule
based models largely outperform the random walk in terms of directional accuracy, the random walk
predicts just 28.83% of the direction of exchange rate movement correctly. For the USDJPY currency
pair the pseudo forecast provides a much higher directional accuracy than that of the random walk. I
conclude that in Japan the pseudo forecast is a better model than the rolling forecast in this country. In
Germany the rolling forecasts all have superior directional accuracy over the random walk model. I
conclude that in Germany the rolling forecast is a better model than that of the pseudo in this country.
These reults provides evidence to confirm my first hypothesis that the Taylor rule inspired forecasting
model beats the random walk model in terms of both RMSE and directional accuracy.
There is also evidence to confirm my second hypothesis that central bank surprise interest rate
movements influence exchange rate movements. For the two currencies that could be modeled with
surprises, the USDGBP and USDDEU, some of the surprise models are superior to that of the no
surprise in terms of RMSE and directional accuracy. For example in Germany the model that
maximizes directional accuracy is the surprise(100%) rolling model. In the UK the model that
maximizes directional accuracy is the surprise(150%) rolling model.
Below are the graphed forecast time series for each of the three currencies. I choose to include the
random walk for each of the currencies and the model that minimizes the RMSE for the respective
currency pair. The blue denotes the in sample forecast and the pink the out-of-sample forecast.
137
USDGBP:
.5
.55
USDGBP
.6
.65
.7
USDGBP No Suprise Model
1995m1
2000m1
2005m1
time
USDGBP
USDGBP_psuedo3
2010m1
USDGBP_insamplepr3
138
2015m1
USDJPY:
80
90
USDJPY
100
110
120
USDJPY No Suprise Model
2002m1
2004m1
2006m1
2008m1 2010m1
time
USDJPY
USDJPY_psuedo3
2012m1
2014m1
2016m1
USDJPY_insamplepr3
80
90
USDGBP
100 110
120
130
USDJPY Random Walk Model
2002m1
2004m1
2006m1
2008m1 2010m1
time
USDJPY
USDJPY_rw
139
2012m1
2014m1
USDJPY_rw
2016m1
USDDEU:
1.4
1.6
USDDEU
1.8
2
2.2
2.4
USDDEU Random Walk Model
1988m1
1990m1
1992m1
1994m1 1996m1
time
USDDEU
USDDEU_rw
1998m1
2000m1
2002m1
USDDEU_rw
1.4
1.6
USDDEU
1.8
2
2.2
2.4
USDGBP Suprise (100%) Model
1988m1
1990m1
1992m1
1994m1 1996m1
time
USDDEU
USDDEU_psuedo2
1998m1
2000m1
USDDEU_insamplepr2
140
2002m1
3.3.2.
Formally testing the hypotheses
The two hypotheses can be formally tested using the Clark and West (2007) (CW) test statistic.
The first hypothesis test takes the null as the random walk model being a superior forecasting model
to that of the Taylor rule fundamental forecasting model. The results of the first hypothesis are below:
Country visà-vis the US
Japan
Model
CW Statistic
Degree of Freedom
P-Value
Psuedo No Surprise
3.14
76
0.00
Japan
Rolling No Surprise
3.15
76
0.00
UK
Psuedo No Surprise
1.72
233
0.04
UK
Rolling No Surprise
-1.80
233
0.04
Germany
Pseudo No Surprise
2.00
140
0.02
Germany
Rolling No Surprise
0.51
140
0.31
The P-Values provide overwhelming evidence that the Taylor rule-forecasting model has higher
predictive power than that of the random walk. For every single one of the models estimated, bar the
Germany rolling no surprise model, the corresponding P-Values are less than 0.05 suggesting that at
the 95% significance level we can reject the null that the random walk is a superior model. In other
words we can accept that the Taylor rule forecasting model is superior to the random walk model.
The second hypothesis test takes the null as the Taylor rule model with the no surprise indicator being
a superior model to that of the Taylor rule model that incorporates a surprise indicator variable. The
results of the second hypothesis test are below:
Country visà-vis the US
UK
Model
CW Statistic
Degree of Freedom
P-Value
Rolling 150% Surprise
0.50
234
0.31
UK
Rolling 100% Surprise
-1.16
234
0.07
UK
5.56
234
0.00
4.78
234
0.00
Germany
Pseudo 150%
Surprise
Pseudo 100%
Surprise
Rolling 150% Surprise
-0.01
141
0.50
Germany
Rolling 100% Surprise
-0.10
141
0.46
Germany
Pseudo 150%
Surprise
Pseudo 100%
Surprise
2.15
141
0.02
0.38
141
0.35
UK
Germany
The P-Values provide some evidence that the Taylor rule model that incorporates a surprise indicator
variable is superior to that of the basic Taylor rule model without the indicator variable. In the UK the
pseudo forecasts incorporating the surprise indicator variable reject the null hypothesis at the 95%
level that the basic Taylor rule model is superior. In Germany the pseudo 150% surprise can be said to
be superior at the 95% level. The other P Values are however insignificant at the 95% level,
consequently providing mixed results.
141
Chapter Four: Conclusion and further research
This paper has provided overwhelming evidence corroborating that a Taylor rule based forecasting
model outperforms the random walk model at the one-month horizon in terms of RMSE and
directional accuracy for the three currencies USDJPY, USDGBP and USDDEU. In addition the CW
test statistic was able to reject the null hypothesis that the random walk model is of superior predictive
ability to that of the Taylor rule forecasting model for all three currency pairs. This finding is
consistent with the wealth of literature that incorporates a Taylor rule model. On these grounds it can
be concluded, contrary to Meese and Rogoff’s view, that economic fundamentals do play an
important role in forecasting exchange rates.
Secondly this paper develops my claim that unexpected surprise interest rate movements by the
central bank can influence exchange rate movements. Although this area of exchange rate forecasting
appears to be reasonably untouched by academics, on the basis of the evidence provided in this paper,
it seems fair to suggest that unexpected interest rate movements do in fact influence exchange rate
movements.
These results suggest a number of promising directions for future research into exchange rate
forecasting. I am not alone in my view that the Taylor rule fundamental based model provides a good
account of exchange rate behavior and I remain optimistic that this is the avenue that should be
explored further. In addition my findings that central banks surprise movements influence the change
in the exchange rate is a relatively unexplored avenue in exchange rate forecasting. Further research
into this area may include applying this model to a much larger number of currency pairs. Moreover it
would be ideal to find a method that could optimize the interest rate differential percentage change
between two countries that identifies as a surprise as I believe this would increase the forecasting
power of this indicator variable.
References:
[Andersen, Bollerslev, Diebold and Vega (2002)] Torben Andersen, Tim Bollerslev, Francis X.
Diebold, Clara Vega. Micro effects of macro announcements: Real-time price discovery in foreign
exchange, The American Economic Review Volume 93, 2002, Pages 38 - 62;
[Campbell and Clarida (1987)] John Campbell, Richard Clarida. The dollar and real interest rates,
Carnegie-Rochester Conference Series on Public Policy Volume 27, 1987, Pages 103 - 139;
[Clark and West (2007)] Todd Clark and Kenneth West, Approximately normal tests for equal
predictive accuracy in nested models, Journal of econometrics 138.1 (2007), Pages 291 - 311;
[Dornbusch (1976)] Rudiger Dornbusch. A re-assessment of the relationship between real exchange
rates and real interest rates: 1974-1990, Journal of Monetary Economics Volume 31 Issue 2, 1993,
Pages 165-187;
[Edison and Pauls (1993)] Hali Edison, Dianne Pauls. A re-assessment of the relationship between
real exchange rates and real interest rates: 1974-1990, Journal of Monetary Economics Volume 31
Issue 2, 1993, Pages 165-187;
[Edison and Pauls (1993)] Hali Edison, Dianne Pauls. A re-assessment of the relationship between
real exchange rates and real interest rates: 1974-1990, Journal of Monetary Economics Volume 31
Issue 2, 1993, Pages 165-187;
[Engel and West (2006)] Charles Engel, Kenneth West. Taylor Rules and the Deutschmark: Dollar
real exchange rate, The Journal of Money, Credit and Banking, Volume 38 Issue 5, 2006, Pages
1175-1194;
142
[Evans and Lyons (2007)] Martin Evans, Richard Lyons. Exchange rate fundamentals and order
flow, NBER Working Papers No. 13151;
[Galimberti and Moura (2013)] Jaqueson Galimberti, Marcelo Moura. Taylor Rules and exchange
rate predictability in emerging economies, The Journal of International Money and Finance, Volume
32, 2013, Pages 1008-1031;
[Meese and Rogoff (1983)] Richard A Meese, Kenneth Rogoff. Empirical Exchange Rate Models of
the Seventies, The Journal of International Economics Volume 14, 1983, Pages 2 - 24;
[Molodtsova and Papell (2009)] Tanya Molodtsova, David Papell. Out-of-sample exchange rate
predictability with Taylor Rule fundamentals, The Journal of International Economics Volume 77,
2009, Pages 167 - 180;
OECD (2015), Industrial production (indicator). doi: 10.1787/39121c55-en (Accessed on 22 February
2015)
OECD (2015), Short-term interest rates (indicator). doi: 10.1787/2cc37d77-en (Accessed on 22
February 2015)
OECD (2015), Inflation (CPI) (indicator). doi: 10.1787/eee82e6e-en (Accessed on 22 February 2015)
[Taylor (1993)] John Taylor, Discretion versus policy rules in practice, Carnegie-Rochester
Conference Series on Public Policy, Volume 39, 1993, Pages 195-214;]
143
Poster presentations
Attendees at Explore Econ also enjoyed learning from the following students who used posters and
multimedia files to explain their ideas.
ο‚·
Minguan Chen (2nd year): Targeting one by targeting all, or targeting all by targeting one?
ο‚·
Dong Ju Lee (Final year): Better value for tax
ο‚·
Jia Hui Lee (Final year): Can the rising tide lift all boats?
ο‚·
Mateusz Stalinski (1st year): Why (not) to join Boko Haram? [Winner Best Poster]
ο‚·
Haoyang Zhao (Final year): Rural-urban migration in China: A brief overview
Acknowledgements
With grateful thanks to the staff and students of the Economics Department at University College
London for ensuring a successful student-led conference.
The conference would not have been possible without the support of:
ο‚·
Deutsche Bank
ο‚·
Faculty of Social and Historical Sciences
ο‚·
Frontier Economics
ο‚·
John Kay
We are grateful for their support in helping us encourage students to Explore Econ!
144
Download